Sybase数据库技术，数据库恢复专家

Others are

Between

Exists

Not Between

Not Exists

Not In

Not Like

Is

Is Not

Developer: added font mapping

pb6.0

the following files will have a [FontSubstitution] section:

pbfont60.ini - Wintel (where win.ini goes)

Powersoft Font Preferences - Mac in the preferences folder

.pbfont60.ini - UNIX in the users home directory

A good example of this is using the

Edit

Cut

Copy

Paste

Ideally if you allow the presentation of Cut/Copy/Paste as a Popmenu(), you would want the pulldown off the sheet menu to match. For example, if nothing is on the clipboard, paste can be disabled by using:

// rightmouse event on sheet

m_edit.popmenu()

// some action causes Paste to disable.

Now Paste is disabled on the popmenu but not on the sheet pulldown menu.

2. Put a dw control on window and associate with the new dw object

3. Run the window and change thewidth of one of the columns so that it's no longer visible

4. Save the syntax of thiswindow using dwsyn.dw_1.dwDescribe("datawindow.syntax")

5. Create a new datawindow based on this syntax dw_2.dwCreate(dwsyn,errors)

It's done two things... (and this isn't consistant behavior.)

1. It over laps the header Text items... (this is almost expected after looking at the old and new syntax.)

2. Sometimes it will switch the order the columns appear in. (I had another example where the columns did show up in the correct order but in the pbl for this issue it switches them.

Cannot recreate during 6.0 beta.

When I tabbed onto the sle there was a "miniscule" movement in the value in the dddw. (when the person submitting this said it moves "ever so slightly" - they meant "ever so slightly") - butit does move when the dddw is editable and it doesn't move when the dddw is non-editable. (I set the priorities as 5 because I think it is such a slight movement that I can't see it causing a problem - but I don't know what goes on under the covers thatcould cause the movement so I am leaving it open since it does indeed move.)

PER DEVELOPMENT:

Works as designed, in edit mode we leave room for the caret.

Cannot recreate using 5.0.04 and 6.0

Want to be able to dynamically populate DW with a graph as well.

Wants to be able to changes the style of the DW dynamically from grid to graph.

He is able to switch it dynamically from tabular --- freeform ---- grid, but when try to get to graph style it comes up without data. He does the X and Y co-ordinates the vertical one shows 0---5---10 and the horizontal comes up as 0.

PER DEVELOPMENT:

XDB driver is no longer supported.

This could happen anytime a user object named UO_<n> is being placed on a window where <n> is the next number for a visual user object control assignment since the painter just increments (uo_1, uo_2, etc).

This works in all versions of 6.0. The patch must be brought forward.

Workaround: The problem isn't the Visual User Object but the collision between the UO_1 global user object name and the fact that the 1st user object inserted (by default) onto a window takes the name "uo_1" (or uo_<n>). This is what causes the GPF. You can:
1. Change the default for user objects prefixes to 'u_'
2. Don't change the default prefix and remember not to name global objects as UO_1, UO_2, etc, and the GPF will not happen.

Developer: Cannot reproduce in 6.0

CREATE VIEW a_view AS

select x0.column from "informix".tablename x0;

PER DEVELOPMENT:

This isn't garbage; it's correct syntax pulled from the viewtext of informix.sysviews generated by informix (the x0's are just aliases). Only if you create a view for log only will PB construct generic SQL.

Development: Development has decided this is a no-fix for 6.0.00 but we changed the dialog for retrieval arguments in nested and composite reports to not have "String Array" or "Number Array" as selectable choices.

The IsDate and Date functions successfully work with a range of years between 0000 and 9999.

Literals against a date variable work with values between 0000 and 3000.

The Edit Mask of mm/dd/yyyy accepts year values from 0000 to 2999.

Even if you open another sheet the dddw will be on top.

WORKAROUND:

//In the RESIZE EVENT of the sheet, code the following:

string ret

// Check the window state to see if the user

// has minimized the sheet.

// If the sheet has been minimized, modify

// the dddw column ShowList attribute to No.

// Otherwise, set the ShowList attribute back to Yes.

if this.windowstate = Minimized! then

ret = dw_1.Modify("column_name.DDDW.ShowList=No")

if ret <> "" then

messagebox("ret",ret)

end if

else

ret = dw_1.Modify("column_name.DDDW.ShowList=Yes")

if ret <> "" then

messagebox("ret",ret)

end if

end if

// end script

PER DEVELOPMENT:

The AutoSizeHeight DrawText code is doing exactly what it was written to do -- strip out leading and trailing 'white space' (i.e. tabs and spaces).

Microsoft has a DrawText function in its Windows API, but it had trouble wrapping text in our text boxes. A PB DrawText function was written to wrap the text, but with white space removal. It has been this way for 3 years. Changing it would impact everyone else that is using AutoSizeHeight.

The proper solution is to add more flags to the AutoSizeHeight check-box in the DataWindow painter property sheet for columns. The user would have the option of stripping out tabs and spaces, or retaining them, as entered.

This has been submitted as an ENHANCEMENT REQUEST.

Workaround: Use a tabular or Freeform dw along with Newspaper columns, Nest the required child and simply make sure the measurement of each row fits on the appropriate Label.

The datawindow was invented when there was no such thing as a desktop color
printer. So, to send out reports with a shaded gray background, most of the time it would make the report unreadable and be a waste of toner.

The workaround is to set a rectangle up in the detail band where you want it. Change the color of the rectangle to whatever you want and make sure you set "Send to Back". You can either:

- make the entire detail band background grey by doing nothing but setting the rectangle color to ButtonFace.
- make certain rows in the detail band background grey (tabular) if you size the rectangle correctly to cover just the column itself and not the entire detail band and put an expression for the background color like:
if(emp_id < 200,255,27775643)

cannot reproduce with pb60 build 152

PER DEVELOPMENT:

I can see how it would be confusing but with a complete column selected (and highlighted) determining and indicating focus is problematic. The sensible thing is to turn off column select when they have done the insertrow and setcolumn.

The customer would like the code to be able to determine whether the border attribute was set to ON, and in that case, override the Autosize Height behavior.

This means the horizontal scrolling still thinks the column is there and if printed, the printout still prints the blank page for the missing columns.

NOTE: This happens in the datawindow painter as well. If you remove columns that go beyond a horizontal page, you will still see horizontal scrolling to the area where the columns were removed and printing will still print out a blank page - just like runtime. If you close and reopen the datawindow, it will recalculate the width correctly.

April 1999 - Development
This works as designed -- the DataWindow Engine only recalculates report widths for grids/crosstabs and nested reports. Removing columns and fields from other DWs does NOT trigger a recalculation of the overall report width. To effect ALL reports would require a major enhancement to the DataWindow Engine.
(This was not a change in behavior in any version - it's consistent - and goes back to the original datawindow architecture)

This issue also applies to MS SQLServer 6.x (MSS). If you use a MS SQLServer 6.x db as the source db, to any other destination db (including a MS SQLServer 6.x db), varchar columns get converted to character columns when you go from the SQL painter to design mode.

Could not recreate in build 5.0.04_61 or 6.0 Beta build 223

PER DEVELOPMENT:

You cannot specify two types of edit styles for a column. Customer wants BOTH editstyle DDDW and editstyle editmask on a column and we cannot do this.

This works fine in 5.0 and 6.0.

NOTE: This won't be fixed in 4.0.

Both NT and 95 seem to correctly respect the case of the filename as it appears in the filesystem. This is the desired behavior. No changes made.

Solution: Implemented in 6.0

Developer: Added code to check if current main library conatins any objects that would be overwritten by doing a gen-app. Presents dialog box with warning if overwrite would occur. In 5.1 builds >= 105. pbapl051.dll

The enhancement would be when printing, if a portrait orientation of a nested report is followed by a landscape orientation, and 'start on new page' is checked, honor the orientation of the nested report as specified in the "print specifications".

Works OK in 5.0

Delta Private Desc: Idle() coded in the open event for the application but while typing in data on thedw the idle event is fired even though typing continuously for a time > idle().. This behavior only happens when the edit control hasn't moved and no mouse activity.

StephenD: You have to stay in the edit control (don't move focus or use mouse) and keep typing. This may be related to messages not being sent or swallowed up by the edit control. Used 4.03.15.

NOTE: You need to type VERY SLOWLY (1 char/second) in the dw edit control so that 8 seconds will go by (idle event triggered) - while you'restill on the SAME column.

Objects and Controls says: ''When the Idle function has been called from an application object script...with no mouse or keyboard activity.''

Workaround: In EditChanged event of the dw control (since this fires with everycharacter) - turn the idle event off via triggerevent of a custom event on the app object for every character typed. In the ItemFocusChanged event, perhaps trigger an event on the app object to re-execute the idle() function.

Workaround: Use primary datawindow for deletedcount() after RowsDiscard()

There is no direct way in PowerBuilder do convert from Decimal (or Integer) to Hex. Told customer Funcky has conversion functions DEC2HEX and HEX2DEC to do conversions back and forth.

Customer would like to enter an enhancement request (even though Funcky can do it) that would convert from Decimal or Integer to Hex. Another choice would be to have an edit mask that supported Hex.

Fixed in 5.0 and 6.0 builds after 6/17/97

Example: On a menu item on a popup menu, the Enable() and Disable() functions will not enable or disable a menu item which is a sub-menu under a menu item which is disabled.

Edit (disabled)

Insertrow (enabled)

If the Edit menu item is disabled, and the InsertRow menu is enabled at first, then doing an m_edit.m_insertRow.disable() will not disable the InsertRow menu item although m_edit.m_insertrow.enabled = false does work.

Development: This is currently correct behavior, the reSelectRow causes the shared DW to be resynchronized and within the re-sync routine it's current row and column are set to row 1 col 1

This requires palette merging and can't be implemented in 6.0

Developer: Added to pb6.0

WORKAROUND: Issue a Filter() to bubble up the contents of the buffer into the display.

PB6.0 build 245 & PB5.0.04_65 - fixed.

Development: Database errors may occur when defining retrieval arguments in the SQL painter. This is due to the fact that we issue a "PREPARE" on the SQL before going to datawindow design.

Development:Original size is a design-time property. It is not meant to be used dynamically. See the following quote from the help file:

"Note :In the painter, if you use the mouse to resize the control or if you set the Width or Height properties on the Position tab, the Original Size option becomes unchecked.

Scripts

The OriginalSize property cannot be set during execution."

Developer: Made a Win16-specific fix to coerce multi-selected filenames to be consistent with their path component.

PER DEVELOPMENT:

Works as designed, print preview shows the printer dimensions which don't translate exactly to screen units.

PER DEVELOPMENT:

Not reproducable in 5.0.03

As far as the Rows/Column Specifications goes, the DataWindow that uses aliases, the DBName is not reflected properly. It also exports the same way. As for the DataWindow that does not use aliases, the DBName looks fine.

Datawindowchild dwc

dw_1.Getchild(dwc,"columnname")

dwc.print()

only prints the headers of the child report.

C0014 - Undefined variable.

Development: Hi-color images in a 256 color environment will repaint when the palette changes which is usually when the active window changes. Functions as Windows was designed.

From Development: This is not bug. The language cannot distinguish between a reserved word and a function name

Fixed in 7.0

Cursor and selectreturn have the same Icon. In DDPLB and PLB controls.

PER DEVELOPMENT:

Though app and EQ pics are there twice (app pics are actually slightly different), they have diff. names. For all these, once the product is released they must stay in because of backward compatibility. Pics are now added rather than replaced.

Using build 4.119 on Win 3.1 client - Sybase V 10 . Export Syntax to Log feature of Database Painter produces redundant syntax - which errors on execution.

I have a table with a primary key defined for it. If I export that syntax to a log file and later try to reexecute it, it results in an error indicating that the unique index that I am trying to create in the third SQL statement already exist. It appears that defining a primary key results in the definition of a unique indexas well so we don'tneed this addiotnal create.

Here's the DDL in the log:

CREATE TABLE dbo.timesheet

(employee_id smallint NOT NULL,

work_date smalldatetime NOT NULL,

task_id char(5) NOT NULL,

weekday tinyint NOT NULL,

hours decimal(5,2) NULL,

timestamp timestamp NULL);

ALTER TABLE dbo.timesheet

ADD CONSTRAINT prim_idx PRIMARY KEY NONCLUSTERED

(employee_id,

work_date,

task_id);

CREATE UNIQUE NONCLUSTERED INDEX prim_idx <----thisis the redundant SQL.

ON dbo.timesheet

(employee_id ,

work_date ,

task_id );

Dropdownlistbox

Possible related issue: ptrack 142401

** 3/4/96: Work Private Desc: France 9956:

If you choose an item in the listbox, and click on cb_2 then it shows the

text attribute which is displayed in the listbox's editbox. If now you

click on cb_1 the listbox's edit box is blank but ddlb_1.text attribute has

not changed. Just click on cb_2 to see.

If you force a second page (i.e., by adding another nested dw in the detail band, etc.), all of the information in the group trailer appears. NOTE: Autosize height has been turned on for the nested dw's and the detail band.

Per Development:
This is not a bug. This issue is another example of breaking the row-processing rules of the DataWindow Engine. We start with the basic rule that a row must fit on a page. Add, if the row is the last row of a group, then the row plus any associated group trailers must fit on a page. Further add, if the row is the last row of the report, then the row (+ any group trailers) and the summary band must fit on a page. If the report is going to break any of these rules, then the DW must be re-designed to utilize nested reports to display the report data -- remembering that the same rules apply within the nest, as well. In this issue, the nested DW makes the primary row too big to be displayed on the same page as the group trailer. The only solution is to re-design the report to get rid of the groups and instead display a series of 'group' details + 'group' summary nested DWs.

In 6.0, this field is greyed out.

Subsequent movement over buttons was also removing the help bar top.

Cannot recreate during 6.0 beta.

Cannot recreate using 6.0

If you delete and re-add the report(s), everything should work fine. Seems like the painter is a little touchy about adding retrieval arguments to the composite after the nests have been added and associated to existing retrieval arguments.

Per Development:
No bug -- when I check the graph properties, I found that the Data properties had the same items entered for the Category and the Values. The Category on the multi-series graph was set to "xvalue, yvalue", and the Values were set to "xvalue, yvalue" -- should have been "Yvalue, Xvalue"!!! The graph was being given (x,x) instead of (x,y) as its data points, and was doing exactly what it was told to do. The single series graph had its Category set to "xvalue" and its Values set to "xvalue" -- should have been "yvalue". When I corrected the Data properties, the graph displayed correctly.

Developer: Will not be addressed until 7.0

Others:

Project Painter - Design->List Object & Options use the same "O" accelerator key

All Painters the HELP option Help-> Welcome to PB 6.0 & Web express use the same "W".

Datawindow Painter under Edit->Copy & Format share the "O" accelerator key.

Object->Group Box and Page n of n shared the "P" accelerator key.

In the Pipeline Painter Design->Undo All & Custom Design shared the "U" accelerator key.

In the Browser under View->Breakpoints & Source Browser the "B" is the shared accelerator key.

Float a painter toolbar by dragging it into the PB workspace. Then, move cursor over left border of toolbar until it changes to double-arrow. Click to resize toolbar. Instead of resizing, the toolbar sort of drags and leaves bits of pixel garbage on the screen. Right side resize seems to work fine.

In preview mode, the top group header lines are not separated by a line. All other groups are. When you scroll down using either the down arrow or the vertical scroll bar, the top group lines are not separated by a line where all the other groups are.

PER DEVELOPMENT:

The result is obtained locally or remotely. The result always wraps to zero.

long val1

long rs1

val1 = 65536

rs1 = val1 * val1

MessageBox("result of squared signed long value of 65536" ,string(rs1))

PB5.0 sometimes doesn't like it: In script editor if you refer to classes or variables with a name beginning with underscore, Powerbuilder puts the brown rectangle over underscore, but compiles OK, and runs OK.

Could not reproduce in 6.0.00

Although clicking Up and Down ScrollBar controls scroll the 'Table Columns' view,

hitting the Down arrow (on the keyboard) when focus is on the control does not

scroll the view.

THIS IS IMPORTANT FOR THE AUTOMATION EFFORT WHICH RELIES ON THE

PROPER OPERATION OF THE UP AND DOWN KEYS.

Up and Down arrow keys do scroll similar controls correctly on the 'Create Index'

and 'ForeignKey Definition' dialogs.

To reproduce:

In script painter, paste the following code:

PowerObject l_Po

// This won't compile: Incompatible property Update for type powerobject

l_Po.Dynamic Update()

// This compilesfine

l_Po.Dynamic ResetUpdate()

Developer: Cannot reproduce in 6.0

The error message is:

C0144: Multiple routines named 'messagebox' were found that match this set of arguments causing an ambiguous match.

The script for a command button was:

string my_string

my_string = dw_1.object.dept_name[2]

sle_1.text = dw_1.object.dept_name[2]

messagebox("the dept name in row two is:", string(dw_1.object.dept_name[2]))

//compile error on the next line

messagebox("the dept name inrow two is:", dw_1.object.dept_name[2])

messagebox(dw_1.object.dept_name[2], "was the dept name in row 2")

Assigning a string variable to the dept name worked fine, as did assigning the sle text to the dept name. The first messagebox also works,as does the third. But the second one gives the error.

GetFocus ().dynamic Undo ()

Developer: That was a design decision that we purposefully made back in 4.0. PBSYC is our System 10 and System 11 interface, so only PBSYC supports the new datatypes. PBSYT is a SQL Server 4.x interface, so it doesn't know about the numeric datatype.

SOLUTION:

This is not a bug. We perform a commit before every transaction we disconnect at runtime. The long delay is caused by dblib waiting for a return code from the backend...

Preview shows extra series key in Legend only for this data source. Only happens in 16-bit. Red box displays by itself. Blue and green are associated with values in series axis data column.

PER DEVELOPMENT:

User error - if you look at the external file included in the zipped file you'll see that there's an extra line in the data .. i.e. an extra newline character .. thus, there's a piece of null data being added to the series

PER DEVELOPMENT:

This is a quirk of the Windows95 Control Panel. When a change is made to the Regional Settings for Dates, the Registry file is updated correctly, but the WIN.INI file is also updated with the change. However, the Control Panel seems to refuse to put MMM in the sShortDate setting of the [intl] section of WIN.INI -- even though it does write the MMM clause in the Registry. PowerBuilder still references the WIN.INI settings for the format of a short date -- to remain compatible with 16-bit Windows. While Windows95 refuses to write the MMM clause to the WIN.INI file, you can manually set it. An sShortDate setting of 'yyyy MMM dd' in the [intl] section of WIN.INI, will be respected by PowerBuilder, but you have to set it yourself.

This was fixed in 7.0

Developer's comments: This is not a bug. It is working as designed.

Would like an enhancement such that the SAVEAS() function has the option of saving to the PDF format.

Function Example:

decimal lc_null

setnull( lc_null )

return lc_null

Powerscript Example:

decimal lc_value

lc_value = wfc_null(<arguments>)

Workaround: Check for null return before assignment.

PSE Update :
-------------------
This cr has been evaluated by development.
Per Development this is not a bug but user error , when setting an expression in a n-up datawindow if it is dependent on a column value then one has to specify the correct row in detail from which to pick up the column value.

Workaround: The execute syntax for the stored procedure comes back:

execute informix.sp_test ;0

by changing it to

execute informix.sp_test ;0 arg1=:arg1, arg2=:arg2

and adding two retrieval arguments of type "string", it seems to work fine as a workaround but this should build the datawindow correctly up front..

Per Development:

Changes required some re-architecting so it will only be implemented in 6.0

See PB Product Management Support Statement : PowerBuilder Rich Text Support

http://techinfo.sybase.com/css/techinfo.nsf/DocID/ID=47864

Fixed in 5.0.02 final.

WORKAROUND:

Declare the argument as a string then use the DATE() Informix function in the SQL where clause to convert it to a date

Workaround: You can get the function call to work by making sure the functions argument declaration is of type DATETIME when passing in a TIME column or value. There is a similar issue with passing a DATE datatype to a global function taking a DATE datatype. It will also convert the DATE to DATETIME under the covers.

The problem with the workaround is once we fix the bug, you'll be back where you started from (having to fix all your computed fields).

GPF: "PB050 caused a GPF in module PScript.drv @ 0004:00000020"

PScript.drv is version 4.00.950 on 7/11/95 9:50 am

NEC SilentWriter 95 and HP 4 ML Postscript. Drivers are NCSW_951.PPD and HP4ML_V3.PPD.

Solution: Could not reproduce in 5.0.03.19. Fixed by bug 41529.

in the following script fragment, no matter what is entered fro em_1.text, Id_mod never equals 0. This worked under 4.0

dec Id_test, Id_mod

// Get the value that was entered in the em

Id_test = dec(em_1.text)

// round to 2 decimals before mod-ding

Id_test = round(Id_test,2)

// Look at Id_mod using the debugger -- it never equals zero.

Id_mod=mod(Id_test,.05)

Id_mod=mod(Id_test+.01,.05)

Id_mod=mod(Id_test+.02,.05)

Id_mod=mod(Id_test+.03,.05)

Id_mod=mod(Id_test+.04,.05)

Development: The following functions are affected: ABS, CEILING, MIN, MAX, MOD.

All of our math functions are prototyped with doubles, and only with doubles, so the compiler casts all arguments to double implicitly--losing the precision of the decimal datatype. Will not be fixed in 5.0, next release instead.

Solution Per Development:

This is a user coding problem that we don't handle very well in 5.0. Is not a problem in 6.0. This problem is not a distributed Powerbuilder problem but a compiler problem.

The problem is the array being passed is a bounded array while the function declaration takes an unbounded array. Since you are passing it by reference this is really not valid to do. This is because since it's by reference you could do whatever you want to the array including change the size. Changing the size of a bounded array is not a good thing. The PB compiler should be catching this but doesn't in 5.0. We have fixed this in 6.0 but have decided that the fix is far too complicated to introduce into a maintenance release. The developer must change either the declaration of the function to take a bounded array or change the array definition to be an unbounded array. Either way when passing by reference the arg must match the definition exactly including if an array is bounded or not.

PER DEVELOPMENT:

This issue involved a grid style datawindow. This is trying to override the featured behavior of the grid style, and will not be implemented.

PER DEVELOPMENT: USER ERROR:

The included pbl copies data from a datasource to a datawindow using a line like

dw_1.object.data = ds_1.object.data

In order for that to work the column specifications of both datawindows MUST MATCH!! In the included PBL the columns appear in the two datawindows in different orders. If one is altered so it has the same columns, IN THE SAME ORDER, as the other then the supplied example works.

PER DEVELOPMENT:

The 1 pixel works fine and will remain in the Sql Painter as is for WinTel. Any issue with the MAC will not be addressed

Cannot recreate in 5.0.03.

Cannot recreate using 6.0

WORKAROUND:

Make the static text box a little bit bigger and then it will print fine.

PER DEVELOPMENT:

Not a bug. This is a printer driver issue.

Works OK in Datawindow painter.

Created a menu. Change and Lock the menu object name. Cascade to a lower level and change something. Return to a prior level and the name has returned to the default, and is now unlocked.

PER DEVELOPMENT:

This is not a bug.

The tab control is modeled after the Microsoft Custom Control tab control.

When the FocusOnButtonDown property set to FALSE the first click on the tab triggers the display of the tabpage and focus is given to the appropriate control on the tabpage. If the same tab is selected again, that tab gains focus as indicated by the focus rectangle. When the FocusOnButtonDown property set to TRUE the fist click on the tab triggers the display of the tabpage and focus is given to the TAB as indicated by the focus rectangle.

Create a function and enter the line: ls_mystring = junk

Save function and get error junk is undefined variable.

From Function painter, go to global variable junk1 and rename junk

Now try to exit the function (should work because global junk1 is now named junk)

Save out of function and you get gpf on pbrte050.dll.

PER DEVELOPMENT:

A user click must occur for any popmenu to close that is already open. This is MS windows functionality,which we have no control over. Additonally, I tested against win95, nt 3.51, w3.11 and nt 4.0 and found the behavior to be the across all platforms.

PER DEVELOPMENT:

The Generation tab of this property sheet affects new generated datawindows and not the one currently open in the painter. This is the intended behavior.

Cannot reproduce during 6.0 beta.

Fixed during 6.0 beta.

PER DEVELOPMENT:

Cannot recreate in 6.0

Memory is leaking slightly but even with memory left, it GPFs eventually.

Product Support Engineering: The behavior exists using 5.0.02. 5.0.03 and up through 6.0 work fine in 16 bit.

If he does not use DPB (Proxy), everything works fine.

Seems to have been fixed in 5.0.02 PBSMI050.DLL

Did a copy/paste of command button to another window and the button works fine. Then go into Declare->User Events of the command button it will display MessageBox: "Duplicate event name".

Named pipes - the error doesn't happen over tcp-ip. It also does not happen in 4.0.06 or 4.0.07 32bit.

Customer has three transaction objects, sqlca and two custom. When he runs the app in development he gets an error while the third transaction is trying to connect: 999 - Request resulted in a bad return code or status but no error message was returned.

Tested on WIN95 and NT, error message is: Unable to connect: SQL Server is unavailable or does not exist. Too many open file handles 10004.

constructor

// Here is the first line;after semi colon

"Override ancestor script"

Exit script painter

after semi colon <---- truncates everything up to semi colon.

NOTE: Will only be fixed in 6.0

PER DEVELOPMENT:

This error does NOT occur with SQLSRV32.DLL 2.50.0126 or in the newer driver that comes with Office 97 2.50.0213. This has been determined not to be a Powerbuilder problem.

PER DEVELOPMENT:

This is definitely a vendor bug. Able to reproduce the error using ODBCTest. ODBCTest reports an error that connectoutstr is getting truncated.

Developer comments: Because of color mapping introduced in 5.0, there were problems with 256 color images. Toolbars still do not support 256 colors images (never have), however, the display has been improved with this fix. The problem with the pictures is that 256 color images scale very poorly when using a 256 video mode. If you set the pictures to full size, they work just fine (this was a difference between the 4.0 and 5.0 pbl).

PER DEVELOPMENT:

NOTE: Works correctly in pre-released 6.0.

Developer: Fixed for 32 bit. Could not be fixed in 5.0. Unfortunately there is a limitation under Windows 3.x (16-bit) for the amount of objects in the common controls. The implementation is based on the usage of standard windows API rectangle structures that use integer members internally. The integer data type range under 16-bit is -32,768 to 32,767. If, for instance, a list-view holds 1840 rows and each row has a 16x16 pixel image and a 2 pixel margin the overall amount of pixels computes to 1840 * (16 + 2) = 33120 => Overflow!

You will always have this limit of about 1800-2000 "viewable" items in 16 bit. Perhaps restructuring the treeview to contain fewer items would help.

Using the keyword "Primary" in conjunction with "Selected" in the syntax to get the values for a column of a dw causes a gpf in pbdwe050.dll.

1) Create a simple, (eg. tabular - employee - psdemodb) dw

2) Declare an instance variable on a window as an array of string (I used string names[100])

3) I tried using the following scripts - :

//names = dw_1.Object.emp_fname.Selected this works

//names = dw_1.Object.emp_fname.Current.Selected this works

//names = dw_1.Object.emp_fname.Original.Selected this works

//names = dw_1.Object.emp_fname.Delete.Selected this works

//names = dw_1.Object.emp_fname.Filter this works

//names = dw_1.Object.emp_fname.Primary.Selected this causes a gpf

//names = dw_1.Object.emp_fname.Primary.Current.Selected this causes a gpf

//names = dw_1.Object.emp_fname.Primary.Original.Selected this causes a gpf

Note: Anytime the keyword Primary is used, you get a gpf in pbdwe050.dll. Both "delete" and "filter" as the buffer type work properly. When you don't specify any buffer type, the data values come from the the primary buffer. (it is the default, you just can't use the word "Primary")

Another note: Seems to be Primary in conjunction with Selected: This works OK:

//names = dw_1.Object.emp_fname.Primary.Current

Workaround: Omit the word Primary; it's the default.

Development: 5.0.04.57 and 6.0 build 226 work fine. The error should be "R1041 Error accessing external object property."

Design behaviour. dw_1.Object.emp_fname.Primary resolves to an array of data, not an object, so adding the selected tag doesn't make any sense. You can't select things from the other buffers anyway.

Cannot recreate 6.0 beta.

Suppose You have a single static text named st_1 inside a windows names

w_move.

Set these position for st_1 as:

st_1.X = 50

st_1.Y = 50

Execute a move (50,50), result is correct of (50, 50)

Then change x's value to 100 and keep y's value the same, execute a move (100,50) and the results are correct (50, 100)

The bug exists if Y is set to a negative number:

st_1.X = 50

st_1.Y = - 50

Reset the X position to 100

st_1X= 100

Execute a move (100,-50), The Y position is changed to -47 even though the Y coord number was never changed via script

Execute another move(100, -50) and the output is a move (100, -43)

The negative Y keeps getting incremented each time a move() is executed until Y finally reached the value of 1.

PER DEVELOPMENT:

The y coordinate no longer changes. Remaining corrdinate differences are do to PBU to Pixel and Pixel to PBU conversions.

Workaround: Use an invisible datawindow control.

Development: This is not a bug. WordBasic is composed of functions and statements. Sometimes the function and the statement have the same name. PowerBuilder assumes you are making a function call unless you use the keyword, AsStatement!. If a function and statement have the same name you will always access the function unless you use the keyword. This problem appeared a little confusing for the German Word situation because PowerBuilder performs a fixup in certain cases. If a user erroneously calls a statement in WordBasic as a function, PB checks the return code for an error message and performs a retry as a statement if appropriate. This fixup only works in English as it is a string compare. Please remember this is a fixup and the user should really be using the keyword, AsStatement!, if they are calling a statement. See the WordBasic documentation on msdn for more info on statements and functions.

WordBasic instructions are composed of statements and functions, which are defined as follows:

A statement carries out an action. For example, opening a file, selecting text, and scrolling through a document are all actions. Virtually every action you can perform in Word has an equivalent WordBasic statement.

A function retrieves information. A function sometimes also performs an action, but the primary purpose of a function is to return information to the macro. A function can determine the character formatting of selected text, whether a document is in outline view, whether a toolbar is displayed, and many other things.

Cannot reproduce during 6.0 beta.

Example: (order_date is a column of type date)

order_date < today()

Today() as a validation expression or computed field expression in the datawindow is the so-called report function that returns a datetime and is not the same as the Powerscript Today() which returns a date. So you're basically comparing a date to a datetime with the example above. The bug would be, that the expression dialog should flag this as an error. "Expecting DATE Expression"

Cannot Recreate with 5.0.04.48 & 6.0 build 214

Cannot reproduce using 5.0.04 or 6.0

PER DEVELOPMENT: USER ERROR:

You simply cannot pass unbounded arrays to bounded arrays by reference. WHen you pass by ref, the types must match.

Steps to recreate:

1. Go into Library Painter (PB 5.0)

2. Select any datawindow

3. Click on Entry, then Print.

4. A blank page prints out.

Workaround:
There is a workaround trick to changing from bar graph to pie and back. Here is an excerpt:

This is still a bug of sorts but graphs (bar to pie) require specific changes with dot notation/Modify due to their nature.

-Series on bar graphs are in the legend, categories on the category axis.
-Series on pie are built into the pie graph labels typically, not the legend. Categories are in the legend for pie.

// Here is some script to change a bar to pie and back.
// Change pie (type 13) to Type 9 "bar chart" "Col3DObj"
dw_1.Object.gr_1.GraphType = 9
dw_1.object.gr_1.legend.dispattr.displayexpression="series"

// Change bar chart (type 9) to type 13 pie
dw_1.Object.gr_1.GraphType = 13
dw_1.object.gr_1.legend.dispattr.displayexpression='category'
// these two lines are key to pie graph labels showing up correctly.
dw_1.object.gr_1.pie.dispattr.backcolor='536870912'
dw_1.object.gr_1.pie.dispattr.textcolor='0'

NOTE: This was fixed in 5.0.03 forward and in 6.0

Developer: This is standard behavior with embedded tabs in a DW edit control and it is in line with the standard Win32 UI.

Although we create the edit controls as "ES_MULTILINE" and use "DrawText" to paint the text under most conditions - and multiline edit controls can support tabs, to do so requires that we do a DT_TABSTOP to set tab positions and DT_EXPANDTABS to paint them properly.

We do not have a function to set tab positions with a DW edit control and it's unlikely that the default of 8 is going to be acceptable. If the column is of "autosizeheight" we use our own DrawText (shWinDrawTxt) to paint the text but we still treat the tabs the same way.

Suggestions: Better to use an OLE column perhaps but embedded tabs are always problematic.

This error occurs when using the command UPDATE STATISTICS; as an embedded sql command. It is only a problem when you issue the UPDATE STATISTICS for the SECOND time in the script code.

From the book for rowstatus: " When a particular status change is not allowed, you can call SetItemStatus more than once to set the row to the desired setting. For example, if you want to set a row with New! status to NotModified!, you can set it first to DataModified! and then to NotModified!." This refers to row statuses but it also applies to column statuses.

Development: This is the correct behavior.
With inserting new rows; the initial row status is NEW! and the column statuses are NOTMODIFIED!

Use setitem() into a column in this new row; the row status is now NEWMODIFIED!, column is DATAMODIFIED!
<< the complaint - Developer now trys to set the column status to NOTMODIFIED! with setitemstatus() and it remains DATAMODIFIED! >>

When a column is set to NOTMODIFIED! with setitemstatus(), PB checks to see if there is "original data" available for the column. If there isn't, we do not touch the status. When we do set a column status to NOTMODIFIED! we want original data to put back in the column value - NOTMODIFIED! implies "undo" in this circumstance and there isn't anything to undo.

<< The workaround is to set the column status to DATAMODIFIED! FIRST, then set the status to NOTMODIFIED!>>

When the developer sets the column status to DATAMODIFIED!, PB saves the contents in an "original column" value, therefore the subsequent NOTMODIFIED! will work. This is one of the quirks of "newly inserted" rows and the fact that there is no 'original buffer' until you go through the chain of DataModified! (just after newly inserted row) to explicit SetitemStatus() of the column to DataModified! (creates original buffer entry) to NotModified! (can do this now since "undo" is allowed with original buffer value).

In this specific case, the only way to set column statuses it to first set them to DataModified! to create the original buffer entry and then to NotModified!.

1. On a window, create a line, changing size, thickness and color. I used blue as color.

2. Duplicate the line, using the RMB popup menu.

3. Size and thickness reproduce, but color does not.

Cannot reproduce during 6.0 beta.

Developer: We actually do not support expressions which require data (ie. and implied row # and column) from other than the primary buffer.

PER DEVELOPMENT:

As designed. The editmask control checks to make sure that an mm has an appropriate value for dd if it is already present thus, if you have 07/31/97 and go to the mm and try to type 11 we will force it to 01 (the first month with 31 days).

Cannot recreate during 6.0 beta

Customer is using:

gr_1.values.labeldispattr.escapement=some value (0,450,900,1800,3600)

gr_1.category.labeldispattr.escapement=some value (0,450,900,1800,3600)

Reproduced using 4.0.06 (16 and 32) and 4.0.07. This does not occur in 5.0.

Cannot recreate this using 6.0 final

datawindowchild dwc

dw_1.getchild('nested_rpt_name',dwc)

dwc.settransobject(sqlca)

dwc.retrieve(61)

5.0 works fine.

PER DEVELOPMENT:

Win95 File caching issue. See bug number 43135 for more details.

Those same characters, if typed into the document after changing the font name from the WFW font DDLB, are fine.

Also, the single quote/apostrophe (') and the double quote (") don't appear, either.

Cannot recreate with 5.0.04.61 and 6.0 build 233

Solution:In 6.0 213, the resizeable checkbox is no longer checkable in the Window painter.

Workaround: Staticbind=1

This is not fixed until PB 8.0.

Workaround: Turn off alignment "Center" or "Right" and use Left or Resize the control or column.

Developer: Fixed in 32 bit. This could not be fixed in 5.0. Unfortunately there is a limitation under Windows 3.x (16-bit) for the amount of objects in the common controls. The implementation is based on the usage of standard windows API rectangle structures that use integer members internally. The integer data type range under 16-bit is -32,768 to 32,767. If, for instance, a list-view holds 1840 rows and each row has a 16x16 pixel image and a 2 pixel margin the overall amount of pixels computes to 1840 * (16 + 2) = 33120 => Overflow!

You will always have this limit of about 1800-2000 "viewable" items in 16 bit. Perhaps restructuring the listview to contain fewer items would help.

Solution Per Development:

When you use VB 4 or 5 and set up a similar test case you get the following results:

1) If you set the SizeMode property to the "Clip" option the fonts look okay, but some of the data is not visible because it does get "clipped.

2) If you set the SizeMode property to the "Stretch" option (which is what PB does), then you get the same results in VB (4 or 5) that you get in PB, the fonts get "scrunched" up so that they're barely legible.

In both scenarios, it's MS Excel that handles the behavior of the fonts.

Development was able to put a "kludge" into the code in order for PB to handle this, but if the "kludge" causes any problems for other customers, the code change will have to be backed out.

PER DEVELOPMENT:

This is not a bug. The reason you can't use the arrow keys on the ActiveX control is because you are tabbing to the ole control container and not the ActiveX. The functionality to be able to tab directly to the ActiveX is an enhancement.

NOTE: Need property on the header and footer band to print only on the FIRST page. This would keep customers from having to do this programmatically.

PER DEVELOPMENT:

This is because 1.17e-38 is not a valid float although the documentation says it is. This is a dup issue of several others. The docs for 6.0 have been changed to reflect this incorrect range.

PER DEVELOPMENT:

Works as designed. Fundamental to the DW design is the "Row must fit on one page" rule

ShowHelp("test.hlp", Topic!, various #'s). 32768 bad, 32767 works.

Solution:

The fix to this bug (changing the definition of ShowHelp( ) in PBTYPDEF would require a migration, so it cannot be introduced in a Maintenance Release. It will be addressed in our next major release, 6.0.

PB 5.03_13 - Focus indicators are leaving dots at its last location when the control is moved + indicators do not go away when another control is selected.

Note: Solaris Only

oleobject oleobj

oleobj = create oleobject

oleobj.ConnectToNewObject("word.basic")

oleobj.FileOpen("c:\test.doc")

WORKAROUND: Either have two DataWindow controls where one can be hidden for printing purposes (using dw_1.modify('DataWindow.print.preview=yes')) and the other for viewing OR before issuing the print for the DataWindow set print preview on and then off after printing. OR modify the size of the columns from the dw painter so all the columns still fit on one page, but large enough to hold all the text.

PER DEVELOPMENT:

Cannot reproduce in 6.0 beta.

2. In decimal col enter 111111.1111.

3. Place insertion point to left of decimal point, Highlight the digit to left, press delete key.

4. Repeat several times (should be 2 times)

5. the edit control displays "-(null).000" in 4.0

Under 5.0.01 32 bit I get: -28127225430599912.000

PER DEVELOPMENT:

Cannot reproduce behavior in 6.0_203

truncate( salary ,2) on the employee table in the Powersoft Demo DB - salary column might show something like:

Salary Truncate

0.29 0.28 // this is incorrect

2.07 2.06 // this is incorrect

0.05 0.05 // this is correct

0.57 0.56 // this is incorrect

2.08 2.08 // this is correct

Workaround: Use a global function.

gf_trucate(salary,<precision>) /* global function */

/* gf_truncate script */

choose case precision

case 2

decimal {2} dec2

dec2= valuepassedin

return dec2

case 3

decimal{3} dec3

dec3=valuepassedin

return dec3

case <n>

case else

Messagebox ("truncate error", "function limit is 8 decimal places")

end choose

Development:

The implementation of representing numercial data as type float (i.e. types float and double) in binary can introduce a degree of imprecision. In performing the "truncate( #, precision)" Powerscript function, the real number (which is "#" above) is involved in a multiplication within the function. The loading of certain real numbers, in this case "2.07", into the floating point register (using the assembly command "fld") in order to perform the multiplication amplifies this imprecision. The represented value loaded into the register in this case is "+2.06999999999999971e+0002", which is dictated by the system floating point representation. A truncation at this point will cause the value to become "2.06". There is not much we can do about this from within PowerBuilder without performing a potentially detrimental side-effect to applications which actually use the trailing decimal places - i.e. scientific data which relies on the proper binary floating-point values. This problem revolves around trying to represent a finite number which cannot be represented exactly in floating point. Other alternatives include integer, long, and decimal, which are designed to represent finite values.

The solution to this problem lies in the hands of the application developer. It is required that the developer use their application-specific knowledge of the data, and perform the appropriate logic (such as "truncate(salary + 0.000001)" to come up with numerical representations that are true within the context of their development projects, and can overcome any rounding problems.

WORKAROUND: The focus problems seem to be related to using None! as a drag icon or Application! if you have no ".ico" file defined for the application object (which is sort of the same as None!). Once you use a real Icon (Information!, or an Application icon file that exists), it works fine.

windowname.title += "string"

Developer: All of this is true, but we are not exposing the flag byte in a maintenance release (would cause compatibility issues). However, the programmer can use the WINAPI function GetKeyState (or GetAsyncKeyState depending) with the virtual keys VK_LBUTTON and VK_RBUTTON to get the state of the buttons. This is a good workaround for 5.0 maintenance.

Cannot recreate with 6.0.

Cannot reproduce 6.0 Beta. Will not be addressed for 5.0.

If you code Destroy Message.Powerobjectparm at the end of the open event of the response window, the problem is solved

PER DEVELOPMENT:

Fixed in 5.0.03. Not applicable to 6.0.

You will see that the path is cut off at the end of the first line or if there was a space in the directory name, the rest of the file path goes to the next line, cutting off any text that doesn't fit.

Using 10.0.03 of Open Client.

You must close the OLE server to have the changes updated in the datawindow.

Developer: There is a function nesting limit of about 58 calls under win16 due to OS limitations (1453 under win32) . The FindItem() function is called recursively in the object function and is therefore hitting this limit. The customer should change their code slightly or redo the treeview structure to have fewer elements..

Product Support Engineering: Fixed with other similar Identity problems either in the specific database driver or in DWE.

Used 16 bit Powerbuilder 5.0.01 and 5.0.02.11 under WFW 3.11. Broken in 5.0.02.10.

There is a related issue with multiple updates when there is an error in one of the update statements.

PER DEVELOPMENT:

A type must be defined prior to use. The order of definition is preserved in the pbl. This restriction prevents 2 structures from being defined in terms of 1 another. Same issue in other languages like C.

Workaround : Export & reorder the structures.

Solution: If your application connections to the database through ODBC and only sets the dbms + dbparm fields, then you are not going to be able to take advantage of transaction pooling.

The following is bad:

sqlca.dbms = "ODBC"

sqlca.DBPARM = "ConnectString='DSN=NetImpact Dynamo Demo;UID=dba;PWD=sql;'"

connect using sqlca; // it will never get a hit in the transaction pool cache

The following is good:

sqlca.dbms = "ODBC"

sqlca.userid = "dba"

sqlca.dbpass = "sql"

sqlca.DBPARM = "ConnectString='DSN=NetImpact Dynamo Demo;UID=dba;PWD=sql;'"

connect using sqlca; // will get a hit

The deal is that our ODBC interface will suck the UserID + Password fields settings out of the dbparm and put them into our internal connection data structure's userid + password fields (not the transaction object fields).

So by the time that the disconnect does occur on the transaction object that originally didn't have the userid + password fields set, they will be set.

If the next connection attempt comes in without the userid + password fields set, then they will not take the cached database connection because the user id + password fields of the attempted connection must match the cached/pooled connection.

possibly only the dw.

recreated in win3.1, may be windows limitiation ??

StephenD: Reproduced only in Windows 3.11. I saw it only took 78 characters in the script painter sheet titlebar. We can fill the titlebar of a sheet from left-to-right in WFW 3.11 with greater than 78 characters. It does cut off so you can only see 78 characters .

PER DEVELOPMENT:

vendor bug - the same thing happens in 3.1 with the titles for message boxes

User gets a "-310 table x already exists" error despite the fact that they are trying to drop and recreate this table.

PER DEVELOPMENT:

retested this with 5.0.03 and 5.0.04 and both worked using the test scenario provided

PER DEVELOPMENT:

Cannot recreate using 6.0.

SOLUTION: This is a duplicate of CR 156272. It is a third party issue. See CR 156272 for details.

July 8, 1998 - Product Support Eng.
There are no further enhancements planned for rich text datawindows so this will probably not be fixed.

per development:

cannot recreate in 5.0.04.

You must reset the return type to (none) each time you edit the event script or it will be saved incorrectly.

NOTE: Will only be fixed in 6.0

Per Development:

This works in 5.0.0x and 6.0.0x. You sometimes have to restart PB to pickup locale changes.

Developer: Expected Behaviour. The deletetitems attribute is for the DELETE KEY only.

codegen compilation error,see file c:\temp\cgen_compile_error.log

"syntax error, declaration statement follows an if statement"

Here is the line of C code generated causing the problem.

_boolTemp0 = err NO_TYPE left { _0bP_this, &_obdataARRAY[0 ] } == 8;

Offending line of Powerscript: (la_local is a variable of type any )

if la_local.typeof() = window! then

/* some code */

end if

Workaround: Use Choose...Case statement to take the TypeOf() out of the IF...THEN..

choose case la_local.typeof()

case window!

messagebox ('window','window')

end choose

Development: Any datatype no longer supports function calls. Closed a loophole in the compiler. Now gives compiler error - actually 2 for the 2 cases.

NOTE: Will not be fixed in 5.0. Needs re-architecting.

This bug affects both 4.0 and 5.0 Powerbuilder. 4.0.06 should use WH1, 5.0 should use WJ2 of the 16 bit Informix Client software.

NOTE: This will be fixed in 6.0 only.

PER DEVELOPMENT: left-arrow and right-arrow function the samve as shift-tab and tab. This is by design and will not be changed.

Development: Broken up to 5.0.02. 5.0.03, 5.0.04 and 6.0 Release. Appears to have been fixed by other fixes.

The workaround of using an invisible datawindow control(s) for each datawindow defeats the purpose of using a datastore to begin with

string s_buffer

s_buffer = "1234"

s_buffer += string(fs_override, "@@@@@@@@")

Workaround:

In the user object function create a temp string variable and perform the string function addition to it. Then assign the value of the temp variable to arg1. For example, replace the existing code in the user function with this code:

string temp

temp="1234"

temp += string("1234","@@@@@@@@@@")

arg1 = temp

return 0

Now 32 bit machine code will compile and run.

=============================

WORKAROUND:

Use a dummy date column in the datawindow result set and dwFindGroupChange function to loop through the datawindow, then getitemdate of the first row in each group, SetItem(date) to the dummy column.

SOLUTION: replace the embedded single quote with 3 tildes and a quote. For example to search for "O'Connor" parse the search string, and modify it to pass "O~~~'Connor" to the find function.

This is probably not a bug, but is being routed to development for forwarding to documentation to clarify the function's behavior, and how it uses all those tildes preceding the quote.

Works OK with pbds, or with 32-bit machine code.

PER DEVELOPMENT:

This is a 16 bit limitation caused by the OS. The call to LoadLibrary is failing. A simple workaround is to rename the .exe. This will only work if you gen everything to use dlls.

Development: Made it so you only need to check the 1 box.

ex: integer_variable = 1 + 2 + this.ic_max

//ic_max is a constant. Declaration: Constant int ic_max=10

Development: Cannot reproduce GPF in PB6 with sqlsrv32.dll 2.65.0240

Development: same as fixes for other null problems. PBCMP050.DLL 5.0.03.06. You must compile with 5.0.03 or higher. It's not enough to just use 5.0.03 DLLs with an executable compiled with 5.0.02, it will still GPF.

If a modify forces a re- retrieval of crosstab dw, the formatting of the DW is lost.

When the user issues a reretrieve of data using a modify statement on the crosstab the format ( look ) of the datawindow changes

dw_1.dwModify("datawindow.crosstab.values='sum(units for crosstab)'") to

dw_1.dwModify("datawindow.crosstab.values='sum(profits for crosstab)'")

In 3.x, if you did a modify, the formatting was preserved. in 4.0.x and 5.0.x, the formatting is lost when you do the modify. Even though the re-retrieve problem introduced in 4.0.x has been solved, the formatting issue is still a problem. A really ugly workaround would be to write a function to re-apply all the formatting, but this could get complicated really quickly...

Workaround: If you change the nvarchar to varchar, the table will pipe.

In 4.0, char datatype is sent to Informix as varchar, and pipeline works.

Development: No specific fix. Fixed with other problems with painting nests and retrieval arguments.

PER DEVELOPMENT:

If "data" is not immediately followed by "[5]", the code assumes that we're dealing with an array; the fact that we eventually get to "[5]" (i.e., ...data.primary[5]) says that this is an array with only one value. The doc is correct in saying that both "dw_1.object.data[5]" and "dw_1.object.data.primary[5]" access current data for row 5. I suppose the Usage description on page 160 could be beefed up to say "when row number is omitted or not immediately after the data keyword" (or some such thing).

Workaround: The float datatype in Informix is being mapped to "real" in the datawindow. Change to "number" and the problem will go away. Users will old datawindows can either rebuild them from scratch (which will remap the "float" datatype correctly to "number") or delete and readd a column to force Powerbuilder to clean this up.

Per development:
No bug -- code is working as designed. The anchor-point for the rotated-text has always been tied to the text-alignment setting of the field. We can't change the algorithm for a patch release. This would have to be an enhancement to the text-rotation feature to add a new Anchor Point property for the Escapement property. Changing the way it works now without adding a new property, would affect reports already in production at the client sites.

Fixed in 6.0. Will not be fixed in 5.0.

Resolved as a result of fix for issue #382233, Bug #41327.

FOR i=1 to 10 /*johwdfjhdjvbscjcsdcojhsdc

pojwdpcojwdpcjwscdpijcoklpkj*/

NEXT

Workaround: Reformat the multiline comment.

also contains a nested report of type graph. When a graph does not fit on the page, the height of the graph gets compressed in height, but part of it gets printed on the next page, overwritting part of another nested report.

"error c0101 referenced object (nvo name), is out of date, must be converted".

Using strange construct in instance variable section containing continuation characters and embedded function calls.

// Calculated Constants

//

int s_DOS_max_filename_length = &

s_DOS_file_prefix_length + &

s_DOS_file_suffix_length + &

Len (s_DOS_filename_separator)

str_test mydata

FOR ii=1 TO 5000

mydata=ds.Object.Data[ii]

NEXT

vcf1net.cnt

vcf1net.hlp

vcfiu.cnt

vcfiu.hlp.

There are from the visual components.

When you paint the WHERE clause in Oracle (using quotes or delimitidentifier='yes'), you get the column list for the where clause in UPPER case automatically.

''DEPARTMENT''.''DEPT_ID'' = :dept_id

SELECT ''DEPARTMENT''.''DEPT_ID'',
''DEPARTMENT''.''DEPT_NAME'',
''DEPARTMENT''.''DEPT_HEAD_ID''
FROM ''DEPARTMENT''
WHERE department.dept_id = 100; << should be uppercase and quotes. This causes an Oracle error.

This is important if the developer is going multi-platform and wants to use just one datawindow.

DUPLICATE BUG: 41077

This is different than having a graph object on the datawindow. Bug 43752.

Getting an access violation when assigning a row from a datawindow to a structure using dot notation where an integer column of this datawindow row has been set to null. In Win95, the GPF is in PBRTE050.DLL Also occurs for decimal and long datatypes.

Development: Cannot reproduce in later versions of 5.0.04 and 6.0

i.e. any my_array[]

my_array = dw_1.object.computed_field.Original

This is documented on page 152 of the Datawindow Reference Guide.

See PB Product Management Support Statement : PowerBuilder Rich Text Support

http://techinfo.sybase.com/css/techinfo.nsf/DocID/ID=47864

ot_get_data_any 'has not been declared'. The script would look something like:

any current_object

ls_CurrentTabPgNumber = MID(current_object.ClassName(),2 ,1)

Workaround: Place the return of any_variable.classname() into a string variable and use the string variable with Mid().

Cannot recreate during 6.0 beta.

Per Development:
Not a bug. The DW Engine only alters the report width for invisible objects when the presentation style is grid, or the right-most object is a nested DW. It has always been this way. It cannot be changed as release patch, as it could affect hundreds of established reports already in use in the client base. Any change would have to be implemented as an enhancement that would only affect a report if the user specifically requests it.

Solution:Cannot recreate in 6.0 233. Will not be fixed for 5.0.

Select Error: Data-conversion resulted in overflow.

The trace shows the BIND goes from len=8 to len=2 causing the overflow. Found in SYB going against Microsoft SQL Server 4.x

Per development:

The ALT key cannot be captured. This is expected behavior.

Numpad '.' works in 6.0. Fixed numpad '.' in 5.0.05.

He is trying to trigger a command button from the connectionbegin . This may be behaviour or bug. Wont work in any version of Powerbuilder.

Workaround: Set lines in dropdown to zero and then 8 lines will be displayed.

Work-Around: Place the exe on a local drive or use WinSock driver instead.

Additional testing showed that there was more to the problem than just testing the 3 months of Oct, Nov, and Dec. My testing showed that whenever the two values entered for month and year were less than 13, i.e. were valid numbers for a month, the editmask simply interprets the incoming string as already being in the dd/mm/yy format and does no checking. If the day number entered was greater than 12, i.e. not a valid month, then further checking was done and the mm/dd/yy format of the string entered was successfully rearranged into the dd/mm/yy format of the editmask control. More details than you probably want follow.

Date editmask. Customer wants users to enter a date in an sle in one format (mm/dd/yy), but display in an editmask in a different format (dd/mm/yy). The editmask does not always correctly interpret the entered date. Details follow, but a summary is that we only use the machine system date setting under certain circumstances when interpreting a date passed to an editmask as a string, and the user would like us to always use the system date setting for interpretation.

**Note that this issue deals with US English settings, and the problem and workarounds must be dealt with in that context. International system date settings would be handled differently within the users code.

The code example for this issue included an editmask control of type date, with an editmask format of dd/mm/yy. The user entered a date in the format mm/dd/yy in a single line edit control. Then the date function was used to convert the entered string to a date, and the editmask.text value is set to the string of that date

(editmask.text = string(date_var)) Using different format strings on the string conversion function had no effect on the results.

*Per Development*: This is due to function internals, and will be treated as an enhancement request rather than as a bug. The following editmask formats work fine, so unless/until this is fixed, use them instead:

dd/mmm/yy

dd/mm/yyyy

mm/dd/yy

WORKAROUND:

Here is the existing customer code. There is an editcontrol em_1. There is a singleline edit control, sle_1. The system date format under regional settings is "mm/dd/yy". The editmask format setting is "dd/mm/yy". The application wants users to enter a date in the sle in the format "mm/dd/yy" and then have the editmask display in the format "dd/mm/yy"

////CODE THAT CAUSES A PROBLEM, in clicked event of a command button:

date ld_date

ld_date = date(sle_1.text)

em_1.text = string(ld_date)

//WORKAROUND IS TO MODIFY CODE AS FOLLOWS:

date ld_date

ld_date = date(sle_1.text)

em_1.text = string(ld_date,"mm/dd/yyyy")

////////////////////

The reason for this is that the editmask control is being passed a string, not a date. It assumes that the string is in the format expected by the editmask definition, unless the

length of the string and the editmask format do not match. Then it will look at the machine's system date setting in order to determine what the valid date format is, and so determine which 2 digit value is the day and which is the month, and do

a proper conversion to the defined editmask format. Adding the format string with 4 digits for the year to the call to the string function forces a check of the system date format setting, and enables the editmask to correctly interpret which value is the month and which is the day, and then to rearrange them to match the defined mask of the editmask control.

The reason that some date values worked the way the user wanted and some don't was that if a date is interpreted as invalid coming in, it would also force a check of the system date format, and would then be interpreted as the user desired. Whenever the value for the day was 12 or less, ie. a valid number for a month, the program would not work as the user desired. If the value for the day was 13 or greater, ie not a valid value for a month, we would check the system date format setting, and figure out that the month was expected before the day, and then do the "swap" from mm/dd to dd/mm desired by the user.

This is best illustrated by examples:

sle: 01/02/97 goes into the editmask as 01/02/97 since both 01 and 02 are valid month values. The string coming in passes the check for both a valid length, and valid values for a date, so the system setting is never checked, and the string goes in as is even though the mask is defined as dd/mm/yy.

sle: 01/24/97 goes into the editmask as 24/01/97, just as the user wants. This is because although the string passed to the editmask passes the length check, it does not pass the data type check because 24 is not a valid month. So the next check is the system date setting of mm/dd/yy. So now the editmask determines that 01 is the month and 24 is the day, and it can put the values in the the dd/mm/yy editmask correctly.

Adding the string conversion format does the following to the 01/02 example. sle value entered is 01/02/97. But the editmask is passed 01/02/1997 instead of 01/02/97. Now the length of the string and the mask are compared, and since they differ, the system date format is checked. The system date format mm/dd/yy tells the editmask that 01 is the month and 02 is the day, so the editmask correctly sets the values to dd/mm/yy as 02/01/97.

Workaround: Issue a Filter() or Setredraw(True). There are many cases with dot notation and SetItem() where changing the buffer doesn't "bubble up" the change to the display. Issuing a Filter() or Setredraw(true) will normally force the display to repaint and you will see all the rows.

Fixed in 6.0 only.

To reproduce:

1. Run the application associated with this bug #, located in the ptfiles directory.

2. Click "OK" on the MessageBox.

Assertion Failure occurs:

rtStatus != RT_EXEC_SUCCESS

In rtPcodeExec,

file rtevthnd.cpp, line 1290

NOTE:

I believe the assertion failure occurs on the line: my_ds = CREATE datastore.

Looking at this in the debugger causes a hang on this line. Running the app. causes the Assert.

2/13/98: UPDATE:

Issue number 421343 is a dupe of this issue. In that case, the code is in a clicked event of a button. There is an easy workaround. Here is the code that breaks and the workaround code:

CODE THAT BREAKS:

long ll_row

datawindowchild ldwc_1, ldwc_2

ll_row = dw_1.GetChild('rep_1', ldwc_1)

ll_row = dw_1.GetChild('rep_2', ldwc_2)

ldwc_1.SetTransObject(SQLCA)

ldwc_1.Retrieve(100)

ldwc_2.SetTransObject(SQLCA)

ldwc_2.Retrieve(102)

WORKAROUND: MOVE THE 2ND GETCHILD CALL:

long ll_row

datawindowchild ldwc_1, ldwc_2

ll_row = dw_2.GetChild('rep_1', ldwc_1)

ldwc_1.SetTransObject(SQLCA)

ldwc_1.Retrieve(100)

ll_row = dw_2.GetChild('rep_2', ldwc_2)

ldwc_2.SetTransObject(SQLCA)

ldwc_2.Retrieve(102)

"Save of OLE Binary information failed. Object must be re-saved to preserve it"

Run application, open window with datawindow and display data. Click command button to SaveAS() .psr (powersoft report) but the report does not get created and a -1 is returned.

Also cannot do Save Rows As....to a .psr in the dw painter, error: Save Rows As Failed

message displays.

NOT A BUG: You need to run "dos\share.exe" for Win 3.1 to save PSR's. It's some requirement of OLE2 runtime(s). Add c:\dos\share.exe to autoexec.bat file.

See PB Product Management Support Statement : PowerBuilder Rich Text Support

http://techinfo.sybase.com/css/techinfo.nsf/DocID/ID=47864

If you code only dw_1.vscrollbar=true and comment out dw_1.hscrollbar=true, then you only get one vertical scrollbar.

m_dynamic mi_dynamic

mi_dynamic = create m_dynamic

destroy mi_dynamic

Dies on the create.

PER DEVELOPMENT:

Cannot recreate using 6.0

Assigning an Edit style of type DDDW to a column in the dw painter, as soon as you select a name from the NAME ddlb it gpfs:

5.0 Module <unknown>@0000:004e6aa3

5.0.01 PBSHR050.DLL@0137:111257a5

5.0.02 PBSHR050.DLL@0137:11125300

5.0.03_1 PBSHR050.DLL@0137:11125500

Used ODBC driver version 2.50.0121 & 2.65.021 with same error resulting. Tested with native db driver MSS, no gpf.

Has an instantiated nvo that calls a uf that calls another uf. The second uf concatenates a string within a FOR/NEXT loop. If string is larger than 4100 characters it results in memory being drained and system locks up.

Support GIF and/or JPEG images natively.

WORKAROUND:

Explicitly delete all subkeys before deleting the parent key.

The code looks something like:

dw_report.DataObject = "d_crosstab"
dw_report.SetTransObject (sqlca)
dw_report.Retrieve ()

It's almost like it doesn't have time to recover from the dataobject assignment before the retrieve happens. If you use postevent for the retrieve() - outside the open, it retrieves fine. It is also fine if you hardcode the dataobject to the datawindow control and take out the dataobject assignment. Again, this is a 16 bit only problem - also in 6.0.01

If you make the datawindow control invisible (so it doesn't have to paint it) you don't have to post the retrieve() so it's something to do with painting and retrieving.

For InfoMaker users, this is built into PBTYP050.DLL in 5.0. In 6.0, you can edit IMSTYLE6.PBL with Powerbuilder and change the open event of w_pbstyle_report and use Postevent() for the retrieve().

July 8, 1998 - Product Support
This appears to be a duplicate of anther bug fixed recently in Change Request 157123 - fixed in 6.0.01 (6.5)

The Design menu item should be enabled, but the Declare menu item is enabled instead.

The only way to get out of Tab Order mode it to click on the Tab Order toolbar icon. You cannot get out of Tab Order Mode from the menu system.

"Open of uo failed. It has been migrated to current version format but must also be regenerated successfully."

Dupe bug report: page() and pagecount() return incorrect value. Dupe reported on the last page of a tabular dw. Last page of dw gives 'Page 2 of 1', which is typical of problem.

PER DEVELOPMENT: The first issue was fixed earlier this year. The issue with large group trailers is functioning as
designed and will not be changed.

PER DEVELOPMENT:

5.0.04.46 and 6.0 build 206 work fine.

PER DEVELOPMENT:

The strange effects of SetRedraw are the result of the way Windows has implemented it. SetRedraw makes windows temporarily invisible and disabled. If you activate another window during this time, you are causing window creation or movement. You can not move or create windows with SetRedraw off.

Per Development:
Not to be fixed.

Also, when an integer or long array is populated from a similar integer or long array, and some values within the array contain null, the resulting values become zero. String arrays seem to work fine.

Development: Not a bug. Must re-retrieve the datawindow.

The inconsistency can be seen in 3 ways:

1) Describe("object_name.Band") returns 1 off. For example, if the object is

in Header[3], then Describe returns Header[2]

2) Describe("DataWindow.Header.2.Height") returns ! exclamation even though

the previous step reports that the band exists

3) Describe("DataWindow.Bands") will not list Header[2] even though it exists.

The behavior is the same with StaticMode on or off.

In IE 4.0, receive:

"Cannot open the Internet Site http://localhost:80/Webagent.pbw/<app name/<object name>/<function name>" The server returned and invalid response.

Workaround in 5.0:

===============

Use the GET method

See PB Product Management Support Statement : PowerBuilder Rich Text Support

http://techinfo.sybase.com/css/techinfo.nsf/DocID/ID=47864

Workaround: You have to distribute the cursor files with the executable.

Per Development - no bug -- user error. The script for filtering the data had the following lines:
dw_1.Sort ()
dw_1.GroupCalc ()
dw_1.SetFilter (ls_filter)
dw_1.Filter ()
The ordering is incorrect. It should be:
dw_1.SetFilter (ls_filter)
dw_1.Filter ()
dw_1.Sort ()
dw_1.GroupCalc ()
That is, AFTER filtering the rows, THEN re-sort the new set of rows, and re-do the GroupCalcs on the new set of rows.
When I changed the scripts, the data appeared correctly with both filter buttons.
When SetRedraw is turned off, the operation are just flagged as required.

PER DEVELOPMENT:

User error. External routine to take a ref int. If you look in the docs for GetUserName you will see that you need to pass it a ref long since it takes a pointer to a DWORD which is a 32 bit number. Works in dev becuase we are in an interpreter.

Product Support Engineering:
Users are trying to share a datawindow LONG column - PB 5.0 and up datatype (from database) to a NUMBER column in the external. Once you change them both to LONG, it works fine.

Older 4.0 INT columns became ''NUMBER'' in datawindows. 5.0 now uses LONG for an INT database column (we're talking SQL Anywhere here) so in many cases, once something is modified in a ''migrated'' datawindow (from 4.0), the painter automatically changes NUMBER to LONG and this can cause these types of problems when sharing to either older external datawindows using NUMBER or database-driven datawindows that haven't been ''modified'' in 5.0 - modified would mean going to the SQL painter and adding or deleting a column - that forces the painter to re-define the datatypes in the datawindow to 5.0 datatypes...

The new datatypes in 5.0 make this sharedata a ''no can do'' even though this would work fine in 4.0.

PER DEVELOPMENT:

Redraw should not be false with the user is interacting with the control. This causes windows to be moved or created.

11/26/97 - Anand KUmar

5.0.04_68 - Cannot reproduce behavior

Similar behavior abserved in the DataWindow painter.

Developers NOTE: Found that client window cannot initiate more than one conversation simultaneously with the same server window in NT 4.0. It seems to work in WIN95 ( however is dangerous: when user calls close the connection is terminated for both, can't differentiate between the openchannels and/or starthotlinks)

There appears to be some behavioral differences in the windows subsystems between NT and WIN95. Both OpenChannel and/or StartHotlink initiate in the same fashion (WM_DDE_INITIATE). Multiple instances of either/both kinds of calls will break under NT 4.0. This is a Microsoft subsystem problem. The workaround is to manage the connections or use different windows for simultaneous connections, between the client and server.

NOT A POWERBUILDER BUG: This problem is due to changes in the Microsoft subsystems, regarding dde, this problem is outside the powerbuilder environment.

ob_inst_defined(obinst) ln ob_destroy_obinst file obmanip.cpp line 1878.

The same code does not get an assertion failure in 5.0.03.

It is unclear as to which date is the modified date and which is the compiled date for the entries. I'm not sure what the best way is to show this.

Regenerating an object (except datawindows) from the library painter updates BOTH dates. My expectations would have been that the modified date would NOT have changed since I did not 'edit' the object.

Developer: Added compile date to entry properties dialog. No changes made to clarify which date is which in the tree view (that's just the way it works) although having both dates on the prop dialog should help. The issue of the mod date changing as a result of compilation should have already been fixed by an earlier release.

Cannot recreate during 6.0 beta.

NOT A BUG:using the paragraph margin setting requires you to select all the text that you want those margins applied to and then perform the set (this can be done programmatically as well as through the Paragraph Dialog box) .. more typically, you use paragraph margin setting to take effect as your typing in text. I think what you really want to do in this case is just set the print margins which are settable via the printspec tab sheet (available at runtime as well as in painter) - or you can set via the dw-modify syntax at runtime.

PER DEVELOPMENT:

Not a Bug. Coding errors specific to test case sent in to support.

PrintCancel() does not work with RTF DataWindows, BUT PrintCancel(printjob) works OK. Both work OK on other types of DataWindows (Tabular, FreeForm...).

This worked correctly in 5.0.03_01

Cannot reproduce during 6.0 Beta

Cannot reproduce with later 6.0 builds.

PER DEVELOPMENT:

This is a feature request and will be submitted to product management as an enhancement request.

32 bit {pb 5.0.02_11, pb 5.0.03_25, pb 5.0.04_31}

This is a win 95 problem only, no problems with NT 4.0

keywords:: datawindow..reports..alignment of static text labels

problem description::

In the print preview a report style datawindow, the alignment of the first static text label is aligned to the right. In the DW design mode all labels are aligned properly under one another.

Customer said that this does not happen on all the datawindows of this presentation style and there is nothing really different about this DW from the other ones.

Second row inserted will always jump to first row when autoarrange is enabled.

*Workaround: Do not use 'for page' in the expression or replace with 'for all' (the default).

If you place it over the window itself, you don't see it at all (except for the shadowbox, if you've specified that for the border of the udo).

If you place it over another control, you see the part of it that is over the other control. If some of the udo spills over the side of the control, you don't see those parts.

Unable to reproduce with 6.0 build 135

Steps to recreate:

1. Open a new window.

2. Place a single line edit on it.

3. Change it's border to shadowbox.

4. Place a group box on the window.

5. Move the sle so that it's just on the edge of the group box. If you try to put it too far in on the group box, the shadowbox disappears. But if you just have it overlap the edge, and then move it in a little more the second time, I was able to get the shadow to appear within the groupbox.

6. Now, move the sle back out of the groupbox. The shadowbox line stays in the groupbox.

However, this is not a real memory leak. Memory is restored after the string array goes out of scope. But reusing the same array elements causes an out of memory message. Since you are reusing the same array elements again and again, you would expect no continued loss of memory.

Development: This is an Operating System issue. The LoadLibrary call (the call that opens the dll) fails. To prove it is an OS issue. I created a simple C program that loads 60 dlls and it fails in that as well on both win95 and winnt.

Note: Solaris Only

fieldID < no_fields in OB_OBJECT_LOCAL::getFieldL value file c:\pb051\runtime\h\obobjloc.h, line 222.

Verified fixed in 6.0.

Tested on 503_25 - All 12 native drivers are present when installing Infomaker

Workaround: Using a mask of mm/dd/yyyy works fine.

Development: This is a serious problem as the pipeline starts a transaction and the MSS driver tries to open a new connection for writing blobs causing MS SQL Server to hold a page lock on the created table and the second connection hangs.

Workaround:
Until the code in MSS driver is modified to not open a new DB connection for writing BLOBs use AutoCommit true for BOTH source connection and destination connection AND use commit factor = 1 in the pipeline painter. This will prevent pipeline to start a transaction and the MS SQL server will not put a page lock on the created table.

Developers comments:

==================

This issue is purely a configuration issue. The following lists the issues to aid customers in resolving these issues.

Windows 3.x & Windows for Workgroups 3.x

--------------------------------------------------------------

To address the configuration issues for Win 3.x (includes Windows for Workgroups 3.x)

we must examine the AUTOEXEC.BAT and CONFIG.SYS files. These files determine the primary environment which is passed to windows when it starts.

For Win 3.x ensure that:

1) the class builder directory is the first directory in the PATH environment variable.

2) the WATCOM environment variable is set to the path (no filename) where the product was installed (eg: C:\WATC).

3) the EDPATH environment variable is set to the path (no filename) where the watcom editors data files are stored (eg: C:\WATC\EDDAT).

4) the INCLUDE environement variable is set to the path (no filename) where the header files are located (eg: C:\WATC\H;C:\WATC\H\WIN;C:\WATC\H\SYS ). If you specify only the header file root directory (eg: C:\WATC\H ) classbuilder will ASSUME that the additional headers are located in WIN and SYS subdirectories. If you have other compilers installed on your system which use the INCLUDE environment variable ensure that the watcom header files are first in the list.

5) no environment varaible exceeds 127 characters in length. With DOS 6.x the PATH environment variable could be 255 characters in length if set in the CONFIG.SYS.

6) wdebug.386 is in the [386enh] section of the SYSTEM.INI. It should be a fully qualified filename (ie includes the path to the file). This is virtual device driver for the debugger. If it is not present or didn't load and you start the debugger you can hang the system.

7) there is enough memory in the conventional memory area after powerbuilder is running.

Every windows virtual device driver and application consume some of this resouce. If you receive Out of Memory errors and the system resources indicate that there is still memory then you are out of conventional memory. There are utilites which can monitor this resource and let you know when you are running low. Other symptoms included system hangs. This problem is exasperated when using Windows or Windows for Workgroups on networks as the network drivers usually use substancial amounts of the conventional memory area.

any antivirus software is disabled while using classbuilder. Antivirus software interferes with the build process.

Windows NT

------------------

To address the configuration issues for WinNT (windows nt 3.x and 4.0 ) we must examine the registry. Note that modifying the registry may result in unrecoverable system failure. You should always uses the control panel applets to modify the registry whenever possible. There are two registry sections or hives that configure NT. The user specific hive HKEY_CURRENT_USER and the system hive HKEY_LOCAL_SYSTEM. The master command environment is determined as follows:

1) HKEY_LOCAL_MACHINE\CurrentControlSet\Control\Session Manager\Environment

2) HKEY_CURRENT_USER\Environment

3) AUTOEXEC.BAT.

4) HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\CurrentVersion\App Paths

The "master" environment is built from the local machine configuration and then the current users environment settings are applied. This means that the PATH environment variable is extended with the current users path, and duplicated environment variables take on the values from the current users environment. The processing of the AUTOEXEC.BAT file is controled from the registry. The default behaviour is to process it. Any environment variables not set in either of the registry keys are add to "master" environment space, as well the last path statement in the file is appended to the current users path.

To this point we have now built the "master" environment that will be passed to every process created on NT. This is not the end of environment processing however. There is optional information that is stored about applications in the registry. This is information that can only be viewed with the registry editor. This information is stored in the HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\CurrentVersion\App Paths key. Every application is supposed to have an entry here. It consists of the application name and extension (eg: PB60.EXE). The default data value for this key is a fully qualified filename for the executable (eg: C:\Program Files\Powersoft\PB60.EXE). There can be additional keys to modify - if you guessed the environment your right - the environment. What is important here is that the PATH environment variable is extended with the directories in the PATH entry, but other environment variables are REPLACED.

This is relevant since PB launches the IDE from with in its own process space. Thus the "master" environment space will be augmented with the environment information from the App Paths\PB60.EXE key. It is the augmented environment that the IDE will be using and not the environment normally seen in the command prompt.

If you experience problems with the classbuilder for NT, you should ensure that:

1) the classbuilder directories are in the PATH environment variable for either:

a) HKEY_LOCAL_MACHINE\CurrentControlSet\Control\Session Manager\Environment, or

b) HKEY_CURRENT_USER\Environment, or

c) HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\CurrentVersion\App Paths\PB60.EXE

2) the BINNT directory must always come before the BINW directory. There are both applications and dynamic link libraries that have the same name in both directories. It is imperative that the NT versions are picked up before the windows 3.x versions.

3) the WATCOM environment variable is set to the path (no filename) where the product was installed (eg: C:\WATCNT).

4) the EDPATH environment variable is set to the path (no filename) where the watcom editors data files are stored (eg: C:\WATCNT\EDDAT).

5) the INCLUDE environement variable is set to the path (no filename) where the header files are located (eg: C:\WATCNT\H;C:\WATCNT\H\NT;C:\WATCNT\H\WIN;C:\WATCNT\H\SYS ). If you specify only the header file root directory (eg: C:\WATCNT\H ) classbuilder will ASSUME that the additional headers are located in NT, WIN, and SYS subdirectories. If you have other compiler installed on your system which use the INCLUDE environment variable ensure that the watcom header files are first in the list.

6) the lmain.cpp file contains a dlllmain section and not just a libmain section. The file should look like this:

/* This file is generated by PowerBuilder.

* You may modify it in any way you wish but do not remove

* Libmain and WEP. Without them you will be unable to link your DLL.

*/

#include <windows.h>

#include "pbdll.h"

#ifdef __NT__

extern "C" {

int __stdcall DLLMain( DWORD, DWORD reason, DWORD )

{

if( reason == DLL_PROCESS_ATTACH ) {

extern char __WD_Present;

if( __WD_Present ) { // this is a hook for the Watcom debugger.

extern void Int3WithSignature( char __far * );

#pragma aux Int3WithSignature parm caller [] = "int 3" "jmp short L1" 'W' 'V' 'I' 'D' 'E' 'O' "L1:"

Int3WithSignature( "DLL Loaded" );

}

}

return( 1 );

}

};

#else

int PB_EXPORT LibMain( HANDLE, WORD, WORD, LPSTR )

{

return( 1 );

}

int PB_EXPORT WEP( int )

{

return( 1 );

}

#endif // PowerBuilder code, do not remove

If it does not look like this then replace it with this one. The HELLO sample that shipped with the 10.5 ClassBuilder for NT suffers from this problem. This will cause a variety of problems such as stack faults, Access Violations, or a hung processe.

7) any antivirus software is disabled while using classbuilder. Antivirus software interferes with the build process.

Windows 95

------------------

To address the configuration issues for Win95 we must examine both the AUTOEXEC.BAT and the registry. Note that modifying the registry may result in unrecoverable system failure. You should always uses the control panel applets to modify the registry whenever possible. The master command environment is determined by the AUTOEXEC.BAT if it exists or by the io.sys file. This "master" environment will be passed to every process created on 95. This is not the end of environment processing however. There is optional information that is stored about applications in the registry. This is information that can only be viewed with the registry editor. This information is stored in the HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\CurrentVersion\App Paths key. Every application is supposed to have an entry here. It consists of the application name and extension (eg: PB60.EXE). The default value for this key is a fully qualified filename to the executable (eg: C:\Program Files\Powersoft\PB60.EXE). There can be additional keys used to modify - if you guessed the environment your right - the environment. What is important here is that the PATH environment variable is extended with the directories in the PATH entry if it exists, but other environment variables are REPLACED.

This is relevant since PB launches the IDE from with in its own process space. Thus the "master" environment space will be augmented with the environment information from the App Paths\PB60.EXE key. It is the augmented environment that the IDE will be using and not the environment normally seen in the dos prompt.

If you experience problems with the classbuilder for NT, you should ensure that:

1) the classbuilder directories are in the PATH environment variable for either:

a) AUTOEXEC.BAT

b) HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\CurrentVersion\App Paths\PB60.EXE

2) the BINNT directory must always come before the BINW directory. There are both applications and dynamic link libraries that have the same name in both directories. It is imperative that the NT versions are picked up before the windows 3.x versions.

3) the WATCOM environment variable is set to the path (no filename) where the product was installed (eg: C:\WATCNT).

4) the EDPATH environment variable is set to the path (no filename) where the watcom editors data files are stored (eg: C:\WATCNT\EDDAT).

5) the INCLUDE environement variable is set to the path (no filename) where the header files are located (eg: C:\WATCNT\H;C:\WATCNT\H\NT;C:\WATCNT\H\WIN;C:\WATCNT\H\SYS ). If you specify only the header file root directory (eg: C:\WATCNT\H ) classbuilder will ASSUME that the additional headers are located in NT, WIN, and SYS subdirectories. If you have other compiler installed on your system which use the INCLUDE environment variable ensure that the watcom header files are first in the list.

6) the lmain.cpp file contains a dlllmain section and not just a libmain section. The file should look like this:

/* This file is generated by PowerBuilder.

* You may modify it in any way you wish but do not remove

* Libmain and WEP. Without them you will be unable to link your DLL.

*/

#include <windows.h>

#include "pbdll.h"

#ifdef __NT__

extern "C" {

int __stdcall DLLMain( DWORD, DWORD reason, DWORD )

{

if( reason == DLL_PROCESS_ATTACH ) {

extern char __WD_Present;

if( __WD_Present ) { // this is a hook for the Watcom debugger.

extern void Int3WithSignature( char __far * );

#pragma aux Int3WithSignature parm caller [] = "int 3" "jmp short L1" 'W' 'V' 'I' 'D' 'E' 'O' "L1:"

Int3WithSignature( "DLL Loaded" );

}

}

return( 1 );

}

};

#else

int PB_EXPORT LibMain( HANDLE, WORD, WORD, LPSTR )

{

return( 1 );

}

int PB_EXPORT WEP( int )

{

return( 1 );

}

#endif // PowerBuilder code, do not remove

If it does not look like this then replace it with this one. The HELLO sample that shipped with the 10.5 ClassBuilder for NT suffers from this problem. This will cause a variety of problems such as stack faults, Access Violations, or a hung processe.

7) wdebug.386 is in the [386enh] section of the SYSTEM.INI. It should be a fully quallified filename (ie includes the path to the file). This is virtual device driver for the debugger. If it is not present or didn't load and you start the debugger you can hang the system.

any antivirus software is disabled while using classbuilder. Antivirus software interferes with the build process.

1) SLE with more text than fits into sle displays differently on Unix than it does on Intel or Mac. This depends on whether text is set in properties or from script, whether Auto HScroll is checked or unchecked, and whether the sle has focus or not.

2) Even with Auto HScroll unchecked, you can use the Home and End keys to get to the beginning/end of the sle, but you have to press them twice.

3) Previously reported: with Auto HScroll unchecked, you can type more into an sle than will 'fit'. Once there is too much text, the arrow keys move past the displayed text, and you can't see the cursor. On Intel or Mac, arrow keys won't move past displayed text.

4) If the sle text is set in object properties, when you tab or click into the sle, the beginning of the text displays, even if the end was showing when the window opened. If sle text is set from script, when you click or tab to the sle, whatever was showing originally remains showing.

PER DEVELOPMENT:

Change category to enhancement:

Want ability to pass how many decimal places we want in the declaration of a Local External Function in User Object Painter

Probably related to other child window bugs determined to be a Microsoft bug by development.

Possible Workaround:

- Map an event to pbm_lbuttonup for each control put on the child window that has this problem with clicking not setting focus to the control.

- In the event, code the following "This.setfocus()"

RESOLUTION: This is an operating system problem. It has been reported to Microsoft. The reference number for the issue is: SR X99041360189

Datawindow specific problems with titlebars

==================================

For datawindows containing titlebars, you will see the same behavior of not being able to position the cursor in the edit control. this problem is a little harder to workaround:

- Create a custom visual user object with 2 datawindow controls. One with it's titlebar (dw_titlebar) and one without dw_dataobject (the one that will contain the dataobject). Position and size the dw_titlebar (turn titlebar attribute on, make the control 3D raised and the background color GREY) just above dw_dataobject and make it show only the titlebar (this datawindow control will never actually have a dataobject assigned - it's used solely to simulate a titlebar on the second control 'dw_dataobject.'

- Create a custom user event on dw_titlebar mapped to pbm_move (call it "move")

- In the Contructor event of dw_titlebar, put dw_dataobject.width = dw_titlebar.width - 5 // accounts for border.

- In the "move" custom user event on dw_titlebar, place the following in script:

dw_dataobject.x = xpos - 8 // accounting for border again.

dw_dataobject.y = ypos

- Call the custom visual user object uo_dwtitlebar and place one on the window. Now when you pick up the datawindow control and they will move together AND more importantly, you can type into the SLE's on the datawindow in the "dw_dataobject" control. The scripting is a bit more involved, but it works.

uo_dw.dw_dataobject.retrieve() or you could write custom user events or functions so uo_dw.retrieve() will do what you expect.

Application terminated. Error: Error accessing external object property"

Development: Will not be fixed in the 5.0 codebase. Use GetItemXXX functions instead. Fixed in 6.0. 16 and 32 bit problem.

Steps to recreate using the pbl in ptfiles:

1. Open d_rte_with_2_computed in the datawindow painter.

2. Go into preview.

3. Click on the menu item design, then header/footer.

4. Scroll up the footer, so that you see "page 1 of 5".

5. Click on the 'last' button on the toolbar.

It now says 'page 5 of 5'.

6. Now, click on the 'prior' button on the toolbar.

PB freezes. When you hit cntl/alt/delete, it shows it as not responding.

Development: The test case does a RowsCopy into another DW, the rowstatus is newmodified! (as expected) and the columns are all datamodified!. When reporting the status of the columns we are making the ASSUMPTION that if the row is NEW and there is SOME value in the column it is datamodified!

The test then sets the rowstatus to datamodified! ... this actually does nothing to the columns status.

When we report the status of columns now, we make the assumption that if the row is NOT new and there is no ORIGINAL data associated with the column then the column status is notmodified!

Remember, we look at original column values as kind of an undo facility and if there is nothing to undo then the data isn't modified except of course in the above example where a row status of newmodified! forces the column status to be datamodified!.

This is 'intended behavior'

Workaround: Go in and out of Print Preview and notice the dashed line (page break) is now in the correct place.

They have table with column defined as number (table created outside PB). In Oracle, this is floating point with precision 38. In datawindow painter, comes up as float.

Some numbers come up with extra precision at the end, and that column cannot be updated.

PER DEVELOPMENT:

have reproduced the above-described behavior. However, the behavior is correct since ItemPopulate is called before ItemExpanding. Therefore the return value of ItemExpanding can't be used as a means to prevent the tree population. In order to prevent the tree expansion and population both methods have to use the counter ii_countit.

PER DEVELOPMENT:

Functions as designed. This is the purpose of SetRedraw(false)

SetRedraw=FALSE should only be used at the start of a performance sensitive piece of code. A SetRedraw=TRUE should follow the performance sensitive section.

long ItestID

double dtest

lTestId = dw_1.GetItemNumber(1,"test1_id") /* long in datawindow */

dTest = dw_1.GetItemDecimal(1,"amount") /* number in datawindow */

lTestId = dw_1.GetItemNumber(1,"test2_id") /* long in datawindow */

where test1_id, test2_id are null. GPF's on the last line.

1) It doesn't respect definition of decimal fields : the decimal part is lost

2) When Windows decimal separator is comma it saves decimal number with comma but dbase wants a dot.

PER DEVELOPMENT:
All that is read from the Windows setting is the separator character. This has always been the case in PowerBuilder, and we do not want to change this because it could greatly impact existing customer applications. A change request will be logged to update the documentation to reflect what we really do: read the separator character.

This does not happen if DISABLE wordwrap.

Per Development:
No-bug -- the DW design breaks the internal rule that the last row plus any trailing group trailers plus any summary band must fit within the printed page. Any overflow off the end of the page, spills over onto the next page without proper pagination. The text doesn't overlap the header text, due to recent patches, but a line of text could be cut in half. There are many formulas withn the DW Engine, that depend on this rule. This will not change until new DW design rules are implemented in some future PB release.

PER DEVELOPMENT:

Unfortunately this is not a bug. It is true in 4.0 and 5.0 and Panther as well. The problem is that even in print preview we are painting to a different (not a printer) device and the font sizes and spacings can be subtly different. So it would be surprising if they DID wrap exactly the same.

Moreover, in this specific example, we are using the Win API call "TextOut" to write the text and we are sending it the same parameters in preview as in print so, someone downstream from us (GDI or the specific print driver) is making further adjustments before it gets on the paper.

A Datawindow print preview is an approximation of how it will look on the paper and cannot be expected to be exact.

This is behaving somewhat differently:

If number return is returned into an integer or a long, 0 is returned if the stored function is standalone. If the function is part of a package, different incorrect numbers are returned depending on whether integer is returned into an integer or a long.

In both cases, stored function does nothing except return 10.

Enhancement: The ability to set an environment variable or setting in the pb.ini file to redirect the placement of the .h, .pch files to the local drive or some other drive so multiple developers don't get collisions.

WORKAROUND: If you create any machine code executable logged on locally into the actual FILE SERVER machine on the network - the one containing the PB software - the .h and .pch files will be created ONCE. After that, all users accessing the network software remotely to create executables will no longer have this problem.

Per development: This is working as designed. CR was revised to go to documentation instead. If this is correct behavior, we should add a note in the ImportFile documentation.

The only way any subsequent connections can be made is by shuting down/restart the server application.

**This only happens on WIN95

PER DEVELOPMENT:

This is a problem with WIN95 explorer. The rename to a long file name is leaving the DOS file name as the old one. If the different names are all specified from Log=, then everything works properly.

32 bit {pb 5.0.02_11, pb 5.0.03_35, pb 5.0.04_31}

Page breaking on a preview of a datawindow is sporadic. Open datawindow ds_report_annual_airlift_main. Preview dw. At the bottom of the first page see ATL HOW. One line (USMC) appears below and there seems to be enough room at the bottom of the page to continue but the rest is continued on the next page.

I canot event preview the dw in nt 4.0 . The app is sucking up resources big time. Normaly it can take about 2 mins to preview , and every time you preview the dw it takes longer and longer until you run out of memory

Fetch: SQLSTATE 01004 [IBM][CLI Driver] CLI0002W Data Truncated . SQLSTATE 01004.

PER DEVELOPMENT:

microsoft bug - notepad gives same behavior

Is this a bug ? Trying other Windows products, most refresh, some do not. For example, Explorer will refresh the file list when changes are made to directories, but File Manager does not refresh if two windows are open viewing the same directories.

PER DEVELOPMENT:

New driver added to 503 code base SYD it is a thread safe version of SYC. syd - needs to be in the vendors list in the .pb.ini. SYD is used just like SYC except SYD is listed as the DBMS.

WORKAROUND: You can include multiple crosstab datawindows with arguments on a composite, you just can't add one later, so you have to create a new composite that includes the crosstab you're trying to add.

Developer: This one isn't a bug, it's one of a class of "bugs" that we call the "good old messagebox" problems. Basically what is happening is that one event is triggered when a control loses focus and that event causes the control to lose focus which causes the event to be triggered which causes the control to lose focus which causes the event to be triggered which causes the control to lose focus which causes the event to be triggered... ad infinitum, as nauseum.

The real point is this: don't use the messagebox to send messages from focus-sensitive events, rather send text to a status object (sle, console window, etc.)

WORKAROUND:

Editchanged event is triggered for each keystroke in the datawindow. Check if KeyDown(KeyDecimal!) = true to trap numeric keypad decimal key.

Specifically, if you think of the events as functions:

//// the following causes a gpf

dw.pbm_dwnkey(keycode key, unsignedlong keyflags)

{

parent.ue_pbm_dwnkey(key,keyflags);

}

////changing the event to this works around the gpf:

dw.pbm_dwnkey(keycode key, unsignedlong keyflags)

{

keycode lkey

unsignedlong lkeyflags

lkey = key;

lkeyflags = keyflags;

parent.ue_pbm_dwnkey(key,keyflags);

}

The user event ue_pbm_dwnkey is correctly defined to expect these two arguments.

str1[] = dw_1.object.data[1,1,5,4]

Debugger shows the full string being assigned to the first member of the array, other members null. In other words only one character should be held for the first element, but instead you could have about 40 characters held. This should NOT be correct behaviour

If you specify a character array of length 1 then only one character is held. This is correct behaviour.

Works OK in 4.x and 5.x with either ODBC 2.0 or 3.0.

Workaround: (works for some). Filter() or Sort() directive in the losefocus event of the datawindow control to force the datawindow engine to re-evaluate the Protect attribute. NOTE: This generally is a problem on datawindows that are completely protected except for new rows inserted. The developer generally wants to immediately protect the new row when focus moves off the datawindow control.

PER DEVELOPMENT: THIS WILL BE CHANGED TO AN ENHANCEMENT REQUEST:

In the user's app, the protect attribute is being set when tab'bing to the second DW, BUT, you can still mouseclick back to the newly inserted row and still change the data. This is because everything else on the DW is protected and we have not completed the "editChange"/"editUpdate" processing, there is no other editable field to change to until another row is inserted. At that point, the protect attribute is in full force.
Here's the word on what is going on with this one:
(Using the user's app)

If you use insertrow() with the commandbutton, mouseclick on the newly inserted row, there is no current edit on the DW so we set an appropriate edit control window over the column.

Type some data

Click or tab away from the column and click or tab back we have an edit control and it's already there ... basically set the focus and that is that. This is where the customer wants us to have evaluated the protect attribute and disallow access to the column.

Insert another row.

Click on the same column as above, again we have an edit control and it is in the right place so we just set focus.

Click in the newly inserted column, we NOW have to move the edit control, go through editUpdate/editChange, the protect attribute is evaluated and enforced from this point on.

So, we ONLY evaluate and enforce the protect attribute for a gob on a DW if the edit control is moving to a new place.

I would be VERY shy about changing this order of events for fear of screwing up other customers or having a BAD impact on application performance as a result of too often re-evaluating too many expressions.

If you use dot notation and enter an incorrect column name, the datawindow error event does fire, so it is possible to halt error processing and allow user another chance to enter a valid column name.

Customer would like the two methods of accessing the data to handle errors in the same way.

PER DEVELOPMENT:

Fixed in 12/16 libuser patch

Create a custom user object uo_globals. Declare and instance constant integer in the user object. Declare a global variable in the application, uo_globals guo_globals. In the script of a window, reference guo_globals.instance_constant_variable. Close the script painter, respond yes to the "save changes?" prompt. Close the window painter. You may gpf here, or you may need to open the window and script painter, make change, and close both (script and window painters) a few times in order to generate the gpf.

PER DEVELOPMENT: Cannot be fixed in 5.0 or 6.0.

WORKAROUND: Use StaticBind=0.

Workaround would be to add date validation yourself to the modified event of the control.

Per Development:

This is not a bug. A spin control editmask does not have the capability to catch this situation. By comparison, an editmask of mm/dd/yy with 03/31/98 can catch the '04' when typed into the mask because we can guess that you think you intend to input that data. A spin control must allow the '04' because you MIGHT be passing through to '05'. We also have to allow this because you MIGHT spin the 31 to 30. An editmask does not get control when it loses focus so it cannot check the final contents, this must be done in script for an editmask on a window. In a Datawindow however, the DW engine will catch the bad date.

Workaround: Regen window before opening. Fixed in 6.0 only.

idw.Object.col1[1] = idw.Object.col2[1]

will cause a "symbol 'ot_get_data_any' has not been declared' error" during compilation.

Workaround: Cast the right hand side to a variable type other than ANY.

long temp

temp=idw.Object.col2[1]

idw.Object.col1[1] = temp

Workaround: If NOT using PFC: Use GetRow() and use that return value with SetRow() instead of the 'row' argument in the dw control.

PARTIAL WORKAROUND for Copy/Paste: Use the clipboard.

rte_1.SelectText(1,1,0,0)

rte_1.copy()

rte_2.paste()

This is happening when running from jogger in development environment. However, for an extremely simple SDI app (2 main windows opened in open event of app, only other code is Halt Close in close event of first window), it works if I run the app when I first come into PB, but if I open another app and run that app, then go back to the problem one and run it, only the first window closes.

Cannot recreate using 6.0.

Test case has a datawindow where all columns in certain rows are protected.

Row Protected:

1 Y

2 Y

3 Y

4 N

5 N

6 N

7 Y

8 Y

9 Y

The following is coded via command buttons. Enter a row number in the sle, and click on the command buttons in sequence.

Function: Argument: Return:

ScrolltoRow 2 1

SelectRow 2 1

SetRow 2 1

GetRow 4 (first unprotected row)

ScrolltoRow 8 1

SelectRow 8 1

SetRow 8 1

GetRow 0

ScrolltoRow 2 1

SelectRow 2 1

SetRow 2 -1

GetRow 2

ScrolltoRow 8 1

SelectRow 8 1

SetRow 8 -1

GetRow 8

Notice the inconsistent behavior of GetRow and SetRow.

This would direct the user as to what to do next. There is a lot of confusion about the interaction of the Paste SQL icon, the DECLARE pasted into Powerscript from this dialog and the "FETCH CURSOR" syntax drawn from the Instance or Global Variable dialogs.

Using Open Client 11.1

The error you get if you don't click on tabpage_2 is

R0039 - Error accessing external object property.

Development: This is a 6.0 change only. Should also fix the dialog in the datawindow OCX to have the same wide properties for the column list.

Loss of Precision caused after using GetItemNumber () on the results of a fractional-number [ a Decimal(3) field created by division of a decimal (real) (dividend) number (with NO precision defined), by an Integer divisor ]

Below is one example of a WORKAROUND for the loss of precision problem.

=================================================================

// MODIFIED Script for event Clicked for command-button cb_1

// ..... the "MAKE-FAIL" command-button

// Script is in file TEST.PBL in PT418639.ZIP

// ............................................................in \\YIPEE\PTFILES\TEMP

// ... shows just ONE example of a workaround (by Suren Raju)

// ... for PTrack Issue 407694

// --------------------- ============

decimal lc_decimal, lc_result

integer idivisor, ipower, idigits, ifactor, iroundpower, iprecision

dw_1.AcceptText()

lc_decimal = dw_1.GetItemNumber(1, 'decimal')

// For this example-case, we want to divide an integer by with 3 digits,

// ... say we have chosen a divisor of 100,that has 3 decimal digits

// ...that is, the divisor has: idigits = 3

idigits = 3

ipower = idigits -1 //Power to which 10 is raised, to divide the number

idivisor = 10^ ipower // Divisor = 10 raised to power 2 = 100

lc_decimal = lc_decimal / idivisor

dw_1.SetItem(1, 'decimal2', lc_decimal)

dw_1.SetItem(1, 'decimal2', lc_decimal)

lc_result = dw_1.GetItemNumber(1, 'decimal2')

// To neutralize effects of Loss-of-Precision., in the dividend-result

// lc_decimal = lc_decimal /100

// We decide how many decimal places of precision are required, for e.g

// If 2 places precision required,we use 10 raised to power(idigits=)3

// ..multiply the "wrong" dividend created, by 1000,then ROUND, and

// ....then divide again by the same number, namely 1000.

//

// Must ROUND to "ipower" digits

// ..So, we round lc_result to 2 positions(iroundpower =idigits - 1= 2)

iroundpower = ipower

ifactor = 10 ^ idigits // ifactor = 10 ^ 3 = 1000

st_message.text =String(round( (lc_result *ifactor),iroundpower) /ifactor)

// This causes the accurate numerical result to be strored in ..

// .... the result-field: LC_RESULT

// Help Information on the ROUND function

// Round ( x, n )

// Argument Description

// x The number you want to round.

// n The number of decimal places to which you want to round x.

// ....Valid values are 0 through 18.

==============================================================================

NOTE: This works fine in NT 3.51, but fails in Win95.

Cannot recreate using 6.0.

PER DEVELOPMENT: 16 BIT LIMITATION: Under Windows 3.x (16-bit) a limitation for the amount of objects in a tree-view exists because the implementation is based on the usage of standard windows API calls that use integers internally.

1) Open a table that has foreign key on a not-null column (I used employee in PSDEMODB which has a foreign key for department - dept_id).

2) Right click the foreign key and select definition from the popup menu.

3) While on the definition window, click on the SET NULL radio button (the bottom one) to change the referential integrity rule.

4) Click OK.

This causes a database error: Integrity constraint violation: column 'dept_id' in table 'employee' cannot be NULL

After clicking OK, the foreign key disappears.

PER DEVELOPMENT:

There isn't a way to 'fix' this issue without a hefty redesign. The 'right' answer would be to issue these 2 Alter Table statements to 'Edit' a currently defined key regardless of the database we may be connected to in one transaction.

Unfortunately, some databases do not support DDL to be executed within an already existing transaction. And some other databases (or at least older version of them) do support creating a Primary Key that has a Nullable column. (This would be a db design flaw to many people, but the older servers only provided a Primary Key as a Logical constraint, not a True database constraint.)

Basically what this means is that this Logic should be 'pushed down' to the individual drivers and if they supported DDL within a transaction, this issue could be fixed for those DB Servers. (But only those DB Servers)

This issue will be added to the list of items to consider for a possible later release, but it will not be resolved 6.0.

This is OK in 5.0.02.

Most likely this is either a difference between installed fonts on the two different platforms, or operating system behavior differences in creating the controls, but submitted to development as possible bug.

WORKAROUND: Manually change the datawindow column sizes to be wider, or the fonts to be smaller.

In general because of machine os and font differences it is best to create controls slightly larger than necessary on the source machine in order to avoid this problem in the future.

Probably related to bug number 33621: gpf when migrated editmask is too short to display data.

Developer: Cannot be reproduced using PB 6.0 build 149.

Workaround: After calling SetSqlSelect(), reset the value of <dw_control>.object.datawindow.table.updatetable to the the proper value.

Example: Assuming the dot notation operand is NULL

dw_1.Object.Field2[1] = dw_1.Object.Field1[1] + "\" GPFs

dw_1.Object.Field2[1] = "/" + dw_1.Object.Field1[1] works OK

When using the PBCGI050.EXE Web PB component with Netscape Fastrack Server Win 95, 32 bit 5.0.04_7.. IDT. Instead of PBNS2050.DLL, you have to place your IP address in the location field of the PbWeb.ini file instead of "localhost". If you use "localhost" you cannot connect to the Application Server; you get the message "Error: Could not connect to PowerBuilder Server 'pbwebtest'".

Timer(1) does not work in an event that was posted to from the window open event. It returns an error code of -1.

crashes

User clicks on bar of graph, supposed to change fill pattern of selected bar. Works in development and p-code but not in machine code exe.

This is fixed in 5.0.03.

PB 5.0.03_25 Win95; severity: 5

When testing a time or datetime display format, the error messages for invalid hour and minute are inconsistent. The hour error message says: "Hour not between 0 and 23", which is accurate if the word 'between' is taken to be inclusive. The minute error message says: "Minute not between 0 and 60", which is inaccurate since 60 is not a valid value for minutes, but 0 is.

Steps to reproduce:

1) Click on the database painter button in the toolbar

2) If the Select Table window appears, click cancel.

3) Select 'Display Format Maintenance...' from the Design menu

4) Click New

5) Set the Type to time

6) Choose a type format such as h:mm:ss

7) Enter 24:00 as a test value and click test, click OK on error messagebox

Enter 23:60 as a test value and click test, click OK on error messagebox

Although the error message suggests that 60 would be a valid value for the minutes, it is properly rejected as being an invalid time. The same goes for the seconds value.

[Group Header]

<report1>

<report2>

<reportn>

[DETAIL]

NOTE: This assumes there is room for the row containing all the nests to fit on one page.

Note: The original fix has been backed out due to problems reported with sharing violation errors if exes/pbls/dll's where on a Novell network. Please see bug #45791.

*See faxback document 2069 which recommends using a copy of the source PBLs for running and debugging rather than running of the "master" copy of the source control PBLs.

Development: Cannot reproduce in 5.0.04.67.

SAMPLE CODE THAT WORKS:

// Set i to Null and use it as argument to Space() function.

// name is correctly set to NULL as a return from Space()

string name

long i

SetNull(i)

name=Space(i)

//////////////end//////////////

SAMPLE CODE THAT CAUSES A RUNTIME ERROR:

// Len() function should return NULL when passed NULL, according to the documentation.

// So i should be NULL after i = Len(NULL_VALUE) but when Space(i) is called

// you get a runtime error.

string name

long i

SetNull(name)

i = Len(name)

name=Space(i)

WORKAROUND: after i = Len(name) change to i = Len(name)

if IsNull(i) then

SetNull(name)

else

name = Space(i)

end if

Fixed in 6.0. Will not be fixed in 5.0.

Grid DW with a computed field.

Wants Rowfocuschanged event to be fired when customer clicks on the computed field.

Clicking on a non - computed field gets the event to be fired.

Workaround:

It does not matter on the style of the dw.. You can get focus to a computed column. So create a new computed column and put the computed column over the computed field to trick the viewer so as to get focus. Maybe use the send to back attribute or transparent attribute to hide the computed column from view of the user so as he will only see the computed field instead.

There is no other real workaround that you can use.

Application retrieves rows into array of structures which is passed back from the server application. Retrieving 4,000 rows OK, Retrieving 4,500 get the msg below.

The ERROR MSG : "This Program has performed an illegal operation and will be Shutdown." " If the problem persists call the program Vendor "

The arrays are declared unbounded on both the client and server side.

FIX PER DEVELOPMENT:Problem was passing any array of more than 4007 elements (give or take). By passing that may mean from client to server, as the return value from the server, or returned through a reference arg from the server.

The product version should not have displayed in this box and has been removed. The Language has been corrected to show English (United States), or the applicable language depending on the build.

Workaround: Place Resizeable=1 in the exported source of every drawing object (rectangle, line, oval, rounded rectangle) you want resizeable.. then re-import and don't change the setting again via the property dialog.

The SetTransPool Function does not behave as expected when acting on the timeout value specified in the third argument. This featured is being used as part of the WEB PB piece of the Internet Developers Toolkit.

This test app uses the SetTransPool function in the Application open event with the settings (2, 2, 10). Time out behavior such as an error saying "Pool limit exceeded" is expected on the third simultaneous connection after a wait of 10 seconds.

SetTransObject later fails with a -1 return value.

Makes it very difficult since you can't read files off the disk to modify them and update into the database. Word works fine.

CGEN: Compiling 'D:\TEMP\424807.c'

D:\TEMP\424807.c(646) syntax error before '_23NULL'; probable cause: incorrectly spelled type name

D:\TEMP\424807.c(646) syntax error

D:\TEMP\424807.c(655) declaration statement follows an if statement

1. There must be a DDDW edit style with no row

2. You must provide a value the first time

3. The first query must return no row

4. You must clear the value when going back in Query Mode using QueryClear

It's not that changing the default setting does not work, but it seems to be dependent upon the order in which the settings are changed. If in the painter cb_1.default = true and cb_2.default = false, if you change the setting in script, the following does not work:

cb_2.default = true

cb_1.default = false.

However, if you code it this way it works:

cb_1.default = false

cb_2.default = true

It's possible that this is a Windows internal processing restriction, but if so it is not documented in PB documentation.

Solution Per Development:

This works in 6.0, no plans to put a fix in 5.0x.

For example, in a 6-up datawindow, you must have inserted 6 or a multiple of 6 rows, or when you try to tab through then off of the datawindow, you get "stuck" on the last row. This does not occur if all columns defined in the result set are included in the display.

Development: Fixed in 6.0.00 in pbcmp60.dll and pbvm60.dll. Cannot be fixed in 5.0 becaues it requires a regen of the application.

Code Before (which caused a memory leak):

Instance Variable:

integer ii_dw_field [250]

Script:

ii_dw_field = { 1, 2, 3, 4, 5, 6, 7, 8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28 ,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102}

FOR li_count = 1 TO ii_NumberOfFields

i_panelfields[li_count].field = ii_dw_field [li_count]

NEXT

Code After With Workaround (no memory leak):

Instance Variables:

integer ii_dw_field [250]

integer ii_dw_field_x [250] = { 1, 2, 3, 4, 5, 6, 7, 8, 9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102}

Script:

ii_dw_field = ii_dw_field_x

FOR li_count = 1 TO ii_NumberOfFields

i_panelfields[li_count].field = ii_dw_field [li_count]

NEXT

Solution Per Development:

If you have 16Mb of space and you try to build an 8Mb blob by concatentation you will have a problem. Here's why: at some point you have a line like

blob = blob + new_bit

well, towards the end of the loop blob has some size (X) that is getting rather large:

X = X + 32K

Now, in our internal representation of this, we have to build the value which is the result of (blob + new_bit) before we can assign it to blob. So we have this temporary value of size (X+32k), and we have the old value of blob which has size X. As X gets close to 8Mb, the total amount of memory required to process that line in the tight loop gets close to 16Mb--and that's not even dealing with the memory the OS requires, other running apps requires, and the Powerbuilder runtime environment requirees...

The point being: you are actually running out of physical memory in all the cases listed.

DB connection is through PBSYC050. The datawindow source select statement is:

SELECT extract.centre_code,

extract.centre_name,

extract.buscat_description,

SUM(location_gla),

SUM(month_1)/SUM(location_gla),

SUM(month_2)/SUM(location_gla),

....

FROM extract

GROUP BY extract.centre_code,

extract.centre_name

ORDER BY extract.centre_code ASC,

extract.centre_name ASC,

extract.buscat_description ASC

The calculated columns map to a decimal(18).

WORKAROUND: Connect to PBSYC050 using StaticBind=0. GPF is caused because the computed column SUM(month_1)/SUM(location_gla) results in a decimal column with 77 digits of precision. When staticbind is on, we don't allocate enough space for this value and our "fetch" overwrites memory.

This works without a problem when migrated to beta 6.0, build 143.

Syntax:

dwobject ldwo

ldwo = dw_1.object.d_employee[1] :d_employee is the nested report

ldwo.object.emp_id_t.text = 'Red' :emp_id_t is text field header

Same problem arises if you get a File I/O error when trying to clear check out status.

Customer is using Oracle 7.3 driver(PBO73050.DLL) & DBParm=DisableBind=0 (default). If she does a select using a host variable other than decimal (long or integer), and its value is greater than 127, no records are found, and she gets a return code of 100.

This does not happen using PBO72050.DLL

*NOTE: This only happens if you try to bind the variable to the SQL statement. If you set DBPARM=disablebind=1, the correct # of records is returned.

Per Development:

The user is posting a function that passes a parameter (a local variable) by REFERENCE not VALUE. It's passing the parameter by REFERENCE, even though the keyword REFERENCE is not explicitly used because the keyword DYNAMIC is being used, and that's the way it works in PB 5.0x.

This is not something that can be changed in PB 5.0x.

This may be a bug, or just may be that we need to add a check for a file extension or at least document that a file extension is required. Probably not a 16 bit vs. 32 bit difference in Powerbuilder, I think it's the operating system difference that when running on a 32 bit OS, the extension is added without the user having to explicitly type it into the file dialog.

Customer has datawindow with columns id, fname, lname and a computed field called test. On id column, has if expression defined for Format property: if (test = 0,"#","#.#") This causes a gpf when the datawindow is previewed. But you can reference the computed field for other property expressions, such as Visible: if (test=0,1,0) works fine. You can also reference a column instead, for the Format property: if (fname>"A","#","#.#") works fine.

when connecting to sybase server 11

Per development: This could not be reproduced at the time of the bug report.

There were other fixes made in this area which they believe fixed this problem.

Selected an object

Selected Source/Register menu item

Received following message:

SCCAPI - ObjectCycle

Operating System Error 3 occured performing Create File

This does not happen when connect via ObjectCycle connection

Customer is getting ORA-01001 error having to do with invalid cursor. Customer is executing a stored procedure and passing a NULL argument to populate a datawindow. This only happened on number type arguments (works fine with char, varchar, ...).

According to the doc, the selected property is used to select a group of objects: "In the list you designate a group of objects by specifying range of row numbers and a range of objects in the format:startrow/endrow/startobject/endobject" In the example, if you specify startrow/endrow/start-with-column-one/end-with-column-three then column two does not end up being selected (highlighted), only column one and column three.

Reported for using dot notation. Using Modify() instead and gets the exact same results.

WORKAROUND: is to use more than one group, specifying all columns as either a start or end of the group: startrow/endrow/col1/col2;startrow/endrow/col3/col4

There is a bug with Syntax #3 of the Time() function.

Using syntax#3 to set a time to "lt_time = time(04,37,29,695)" results in "04:37:29.000695" (this is the bug).

Using syntax#2 to set a time to "lt_time = time("04:37:29:695")" results in

"04:37:29:695000" which is correct.

PER DEVELOPMENT:

This is design behaviour and has to do with the format of the parameter. When the parm is an integer it represents the number of microseconds, whereas when it as a string it is intended to function as a rational fraction. Ie. 12:12:34:345 means 34.345 second after the 12th minute of the 12th hour, which means 345000 microseconds.

Transferring to documentation group, the documentation for Time(string) should be more explicit about this.

Error: "MKS Interface Error: Invalid Arugement"

**This works in 5.0.04 & 6.0

Issue happens when you add a tab control to the ancestor window and the goto the descendant(child window) and save the changes there you will get the above GPF

File Handling function. Trying to do a file open on an unqualified file. It is coming up as a #1 instead of the actual file handle.

A workaround is to specify the full path name where the test.ini file is. In my case I put in in the pb4 directory or the pwrs\pb5i32 , both of which are specified in my path.

The "fix" is a change in the default behaviour of version 5.

As a general rule we don't put changes to default behaviour in maintenance releases so the problem eill be addressed in PB 6.0

Three ways to bring up a new application:

1) Application Painter, File Open, ...

2) User double clicks on an app object in the Library Painter

3) User selects an app object in Lib Painter and chooses the Entry Edit

development: This is a documented limitation in the datawindow for numeric types. 18 decimal places.

PER DEVELOPMENT:

Fix for 47588 fixed this issue.

The customer inserts an rtf doc into dw-any charac that are tabs or indents show up as transparent. happens under 32 not 16 bit pb.

PER DEVELOPMENT:

Fixed in 6.0 only.

If you click the delete (clear) toolbar button, the button stays down and the delete action is not performed. You cannot delete entries using the preview mode if this is broken.

To reproduce:

1. Create a datawindow

2. Preview the data and then select a row and click the delete button

3. Notice that the button remains depressed and the delete does not happen.

Works if you save the landscape setting in the dataobject and if you use

dw_control.modify("datawindow.print.orientation=1")

August 12, 1997 - Works fine in 5.0.03 final and 6.0. Definitely a problem in 5.0.02

R0034 Error: Incorrect number of parameters calling external object function textrc at line...

Workaround: don't use the exact same name.

"%s already exists in the source management archive. Register this PBL entry anyway.

*** Workaround

==============

Here's the original code it the menu item:

W_TESTLV lw_temp

lw_temp = This.ParentWindow

lw_temp.lv_testlv.EVENT POST ue_rename ()

if you replace it with the following:

ParentWindow.DYNAMIC EVENT POST ue_rename()

Everything works fine.

FOR i=1 TO 50

tab_1.OpenTab(iu_with,"u_with_tab" ,2)

tab_1.closetab(iu_with)

NEXT

j=8 <- dies here at next write to memory after exiting the loop.

This problem does not exist using Beta 6.0, build 155.

ENHANCEMENT REQUEST.

Unable to recreate in build 153

Solution Per Developer:

When using mdi custom, managing the mdi client area becomes the responsibility of the pb developer. Put the script code for the frame resize event in the toolbarmoved event also. There is no resize on the opensheet , but there is a toolbarmoved event.

ole_1.object.navigate("http://www.powersoft.com").

works fine in NT 4.0 but GPFs under Win95.

The GPF is in SHDOCVW.DLL @ 0137:715159ff.

NOTE: This no longer GPFs in 5.0.04.68 or 6.0.00. PBROI050.DLL 5.0.04.68 combined with 5.0.03 solves the problem as well.

Similar to bug 41850.

Code that fails:

int array1[5] = {6,7,21,9,10}

int array2[5]

array1[1] = 1

array1[2] = 2

SetNull(array1[3])

array1[4] = 4

array1[5] = 5

array2 = array1

//This incorrectly sets the values of array2 to {1,2,21,4,5} instead of {1,2,null,4,5}

//The same problem existed for bounded to unbounded and unbounded to unbounded //array assignments.

WORKAROUND:

Instead of array2 = array1, code:

for i = 1 to upperbound(array1)

array2[i] = array1[i]

next

This works fine and preserves all values correctly, including nulls.

: Trying to use MS web browser control OCX. If you put the control on a window, then try to get the OCX properties, the popup menu goes away and nothing else happens.

using the script ' string ls_web

ls_web = "http://www.powersoft.com"

ole_1.object.Navigate(ls_web)' in the doubleclicked event of the OCX will cause nothing to happen. You cannot double click the control

This works in 6.0

Work around is to have a non-key column in the table.

// emp_id is an integer column from sql anywhere

string string_var

SELECT emp_id into :string_var from employee;

will compile since the "emp_id" column and table "employee" exist but the Powerscript editor won't check to see if the "INTO" variables are compatible with the host variable :string_var.

Dynamic SQL has the same type of issue and would be even harder to check. Powerscript leaves it up to the developer to make sure the datatypes between Powerscript and the database match and are compatible.

// Dynamic SQL example

string sqlstring, ls_var

int li_var

sqlstring = "Select emp_id from employee where emp_id = 105"

...

open <cursorname>;

fetch <cursorname> into li_var;

or

fetch <cursorname> into ls_var;

both compile - and both actually work too in Wintel platforms - go figure - no runtime error.

table1.col1

table1.col2

table1.col1_1 <---- these should not be there.

table1.col2_1 <---- these should not be there.

table2.col1

table2.col2

It should be:

table1.col1

table1.col2

table2.col1

table2.col2

NOTE: This is not database specific. Recreated with Sql anywhere.

Select error: SQLSTATE=01003

[IBM][CLI Driver] SQL0000W Statement processing was successful.

or

SQLSTATE 01S01 [IBM][CLI Driver] CLI 0165E Error in Row

The desired object of the previous version should be created - the tip object should not be created.

Customer is trying to assign an array of float variables from an external function. He has declared a bounded array of either double or real and has the same problem:

R0047:Array expected in ANY variable at line ### in Open event.

When he uses an array of strings, this works fine.

DOUBLE FloatArray[2], fl_dec

OleObject obj

obj = CREATE oleobject

obj.ConnectToNewObject("MFCAutoItem.AutoItem1")

FOR i = 1 to 2

FloatArray[i] = 1234567

NEXT

IF (obj.SetFloats(FloatArray)) then

messagebox("successful", "", Question!, OK!)

END IF

FOR i = 1 to 2

FloatArray[i] = 0

NEXT

// Error: Array expected in ANY variable when using DOUBLE

// you must use safearrays when passing arrays using OLE

// automation

FloatArray = obj.GetFloats()

The floatarray(safearray) is of data type float, if it were of string or integer there would be no problem.

32 bit { pb 5.0.02, pb 5.0.03, pb 5.0.04_39 }

Keywords:: RTE, DW , print and print preview

Problem description::

Customer is using an RTE dw. If the inserted text fields are too long the text does not wrap onto the next line. Instead the line has no blank spaces between words!

In print preview you can see the first line "squashed" together. There are no spaces between the words as you would normally see in the dw design mode.

WORKAROUND: Use the same dw expression for the Protect property, and now you cannot tab to the column when it is invisible/protected.

dberrhandle((DBERRHANDLE_PROC)err_handler);

dbmsghandle((DBMSGHANDLE_PROC)msg_handler);

Error under 5.0:

"The instruction at 016a23b5 referenced memory at 00001645, Memory could not be read."

Error under 6.0:

"The instruction at 0x03063f64 referenced memory at 00001645, Memory could not be read."

Select ....... where emp_fname LIKE 'A%' AND emp_fname LIKE '%B'

Workaround: after setitemstatus() of row and primary key to Datamodified!, must do a setitem() of the primary key column to itself (same value). This will generate the DELETE and INSERT.

Solution:

This issue will not be fixed at this time. There is a reasonable workaround and the fix has the potential for breaking a large amount of customers already using this functionality.

1. the datawindow contains a dropdowndatawindow that is editable.

2. there are rows in the filtered and/or delete buffers of the datawindow

3. there is code in the datawindow's editchanged event. Unfortunately, we have many datawindows like this and have not found a workaround yet. The code consisted simply of a comment, no actual programmed activity.

Per Development:
Not to be fixed.

-> Crosstab : clicking on "Data Source" in the PainterBar causes "X server error..." to display on stderr.

Here is the complete error message displayed on standard error :

X server error, Message reported is: BadDrawable (invalid Pixmap or Window parameter)

There is no "fatal" event that occurs after the above error is displayed. This behavior has yet to be tested on Solaris and AIX.

*******************************************************************************

[BSmith] This bug was reported again but in Solaris It is as follows

PANTHER - B103. REGRESSION SOLARIS ONLY When creating a Quick Select DataWindow the following Console Message appears:

X Server Error, Message is: BadDrawable (invalid Pixmap or Window parameter)

Steps to recreate:

1) Create a New Quick Select DataWindow.

NOTE at the current time unable to create a SQL SELECT DataWindow.

As soon as you click the OK button after choosing the Presentation style, the above message appears 4 times.

We currently have a 64k limit for Modify() and 32k limit for Describe() of the "datawindow.table.select" attribute.

Development: For architectural reasons , only fixed in PB 6.0.

PER DEVELOPMENT:

Worked with customer and found subtle mistake in profiles being used. Under more usual circumstances this would have been obvious, but it took some investigating in this situation. No further problems here.

Workaround: Before saving the pipe, edit the definition created by the pipeline painter so the "NULL" checkboxes match the source (Informix 7.x) table for each column being piped.

NOTE: Requires 5.0.04 version of PBIN7050.DLL to use the new functionality of reading the registry "setnet32.exe" entries and using host@dbservername syntax in sqlca.servername.

The Save Rows As... bugs are:

-datetime columns, YEAR TO FRACTION(5) isn't generated.

-integer columns are converted to FLOAT.

-primary keys, the "NOT NULL" attribute is NOT generated.

-decimal(5,2) becomes the default decimal (16,2)

-smallint becomes FLOAT.

-real becomes FLOAT.

-money(8,2) becomes the default decimal (16,2)

-varchar(10) beomes char(10)

Workaround:

1. Object/Export Syntax to Log in the DB painter. This saves the CREATE syntax correctly - unlike the Save Rows As.... Save this syntax without the INSERTS into the PBCatXXX tables.

2. Preview table in DB painter. Save Rows As... SQL.

Take out the CREATE syntax since this is messed up. Cut and Paste the INSERT SQL (for the rows to be inserted) - this syntax is OK.

3. Add the two files together.

Development: Currently this generation does not involve the database drivers. In a future release, we could leverage our "Export Syntax to Log" functionality to generate vendor-specific syntax here as well.

"Title Bar" should be linked to the other three. You can't have the Control Menu, Minimize, Maximize Buttons without first having a title bar on the datawindow control. Notice if you right mouse on any datawindow control, you'll see "Title Bar" checked OFF but the "Minimize", "Maximize" and "Control Menu" are still enabled but they shouldn't be.

If you play around with it enough, it eventually acts correctly but not with a newly-placed datawindow control.

Probably PSReport! should not be allowed as a saveas format for a graph control on a window, but the documentation indicates that this is a valid format.

WORKAROUND: Use a datawindow graph object instead of just a graph on a window.

Development: Microsoft has made huge improvements in their ODBC driver to offer built in networking etc right in the ODBC driver. As a result, the Microsoft ODBC driver is not like traditional drivers (like Intersolv) where the ODBC driver is a layer on top of the native software.

IT IS A MYTH that ODBC must be slower than native protocols. Intersolv's driver is slower since it sits on top of the native SQL Server drivers.

NOTE: PB Desktop does not allow access to the SQL Server ODBC driver.

datetime ldt_1, ldt_2

ldt_1 = Datetime(today(),Time(3,36,12,45))

ldt_2 = Datetime(today(),Time(3,36,12,89))

sle_1.text = string(ldt_1,"mm/dd/yy hh:mm:ss.ffffff")

sle_2.text = string(ldt_2,"mm/dd/yy hh:mm:ss.ffffff")

mle_1.text = "Notice that the only difference is in the microseconds portion of the 2 datetime variables"

IF ldt_1 = ldt_2 THEN

messagebox("equal","If you see this message PB thinks that the 2 datetimes are equal even though they are not. Congratulations you've reproduced the bug")

ELSE

messagebox("not equal","Looks like the version of pb you're using doesn't exibit the bug behavior")

END IF

See also bug 43835

PER DEVELOPMENT:
It is standard behavior for the tab key to to move focus between controls. To enter a tab in an mle, the user should use the Control-Tab
(CTRL-Tab) key combination.

NOTE: There are several other graph object printing bugs. 40131 was a problem with a graph object in the detail band and set to foreground. 42432 was another similar one - both fixed in 5.0.03. This bug is different and wasn't fix until 5.0.04.

Development: Cannot recreate in 6.0 build 156.

The preview works fine. On paper, the first column values of the horizontal pages are shifted to the right. The printer is an HP 4si.

Example:

m_ancestor - ancestor menu

m_decendent - decendent

Opening m_decendent when m_ancestor has a menu item colliding with global menu object m_decendent causes the GPF. It's generally with migrated applications since the painter now checks for this case. Migration and regeneration doesn't catch this. Very hard to track down.

When you use 'USER ID' - in the profile painter as "informix" (owner of the catalog tables) for example, the data pipes fine. The USERID is required ONLY AT RUNTIME for most of the native drivers as a feature that allows runtime overrides for generating prefixed sql.

Workaround: Set USERID in the profile painter to be the owner of the tables. You shouldn't have to do this in development.

PER DEVELOPMENT:

Problem caused due to pb.ini having the read-only attribute set. Fixed after making the file read+write.

PER DEVELOPMENT:

Shadow borders only appear on their parent window. Moving a control over a tab page only moves it over the tab page, it does not make it a part of the page. The tab control and listview are siblings in that situation. Not a bug.

PER DEVELOPMENT:

This is not a bug. It is expected behavior.

The group box, and indeed any control, is not part of the tabpage unless it is created on the tabpage. Note that you cannot drag a control off the tabpage either. This is by design.

Per Development:

Original problem fixed.

Also, unlike NT, the Oracle 7.3 API (OCI) for Sol/HP/AIX does not provide a call to invoke thread-safe mode, therefore no checkbox. We'll have to take Oracle's word that these APIs are thread-safe by default.

There are problems at the moment with windows 95 plus. If the customer uses the above workarounds he is all set.

The combination Win 95 4.00.95.0b and Novell network driver is failing the printing of a dw to a printer queue. If you do one of the following, it works fine:

- use an earlier version of Win 95

- use the Microsoft Novell client instead of the Novell Novell client

- print from PB something else than a dw (a script or an object)

- redirect the queue to a port and print to that port instead of directly to the queue

Customer is trying to display bitmap in a dw. Has a dddw, which has been mapped to several bitmaps from parent dw. The property of the column object has set to Display as Picture, and for the edit style (DropDown Dw), the property has set to Always show List. When list is displayed if any of the items get focus, it shows as the path to the bitmap instead of the picture.

If you do a regular datawindow update and a database error occurs, the DBError event fires and SQLDBCODE & SQLERRTEXT are populated with the correct information.

PER DEVELOPMENT:

Verified NAB in 6.0.0.248 and 5.0.04.67

Please check your islv.ini file living either in the windows directory or dir defined by system variable ISLVINI.

The same problem occurs if the computed field referencing a retrieve argument is in the header.

R0014: Error opening dll library pbtyp050.dll for external function

Workaround: Use the 5.0.02 16 bit PBTYP050.DLL for development. They should be 974k.

The 32 bit versions are 964k.

NOTE: The maintenance release for 5.0.03 on the Web site will have the correction as will any new CD's that are built. Affects customers who received the first set of CDs.

Development: This was fixed with other changes to the runtime.

Workaround: in this situation, if you can change the graph line color to be the same as the color of the category, it doesn't show.

Watching the resource monitor on NT 4.0 you notice that resources move up about 10%, but never fall back down the way it should. It appears that resources are being taken up but never freed.

Using LibraryImport in a distributed server app always fails if their is a client connected and you are trying to import into a pbl that is in the server's library list.

LibraryImport will work with a standalone application and should work in a distributed application . According to development it should work but something is going wrong

From Development (8/20/97)

This is an architectural problem that cannot be fixed until 7.0.

transferring as an enhancement

Developer: Cannot be fixed in 5.0 because it would require a migration

This may be a limitation of the Label datawindow style, but it is not documented. Also, the expressions property sheet for the label datawindow column does not include other properties that can apply to a column, so just the fact that width and height are listed would imply they could be modified. (Example, the Visible property, valid for columns in most datawindow styles, does not appear in the expressions property sheet for a column on a label style datawindow.)

Possible workaround: Use an n-up datawindow presentation style.

Development: Bristol layer fix.

Confusion between checked and 3rd state.

The window control - correctly - has a dark grey background with a slightly darker grey check - like VB 4.0.

WORKAROUND: We're not picking up the default Buttonface color.

For NT 4.0.

-Go into Display settings

-Appearance Tab Folder

-Item: 3D Objects (dropdown)

Notice, the color default will be buttonface. Click Apply and the next time you preview the column, the 3rd state will magically show correctly.

For Win95, it's slightly different. Simply clicking apply doesn't work.

-Go into Display settings

-Appearance Tab Folder

-Item: 3D Objects (dropdown)

-Change buttonface to some other color like Red and "Apply"

-Change Red back to buttonface and you'll see the 3rd state correctly.

Development: For datawindows to be transparently used across databases, developers must set staticbind=0 in their SQLCA.DBPARM (script) or database profile painter. Unfortunately, the default for any connection is staticbind=1 (do not re-bind result set).

No fix is planned. We will try to document this better so users migrating from one database platform to another will be aware of this requirement.

Development: This is not a feature of the datawindow currently. Users could use ole1 blob column or some other way so saved data with embedded control chars can be interpreted corrrectly.

This happens, for example, when an argument to the global function is a user defined object (structure, window) contained in a library not on the library list. (could be an object function too)

The message you'll get when you try to re-open the function:w

Open of function f_anymath failed. Possible Causes:

1. Object has been deleted.

2. Its ancestor has been deleted.

3. An object it references has been deleted

4. Object was created with a newer version of Powerbuilder.

Migrating after adding the missing PBLs containing the missing objects does not fix the problem. If you export the object or global function, you'll see there is no script in it.

Fixed by patch received from Bristol Technologies

Workaround: Use External function calls (The following are 32 bit calls)

Function ulong LoadCursor(ulong hi, ref string str) Library "USER32.dll"

Function ulong SetCursor(ulong hCursor) Library "USER32.dll"

Function ulong SetSystemCursor(ulong hCursor,long ll) Library "USER32.dll"

Function ulong GetCursor( ) Library "USER32.DLL"

Function ulong CopyIcon(ulong hInstance, ulong hIcon) Library "USER32.DLL"

Developer: Fixed by patch recieved from Bristol Technologies.

In 5.0 thru 5.02 when an oracle stored procedure was pasted the datatype (double) of the stored procedure was pasted correctly. In 5.03 the datatype is defaulting to (decimal). If you have coded a variable as a double and passing it to the RPC call it will now give an error "referenced arguments do not match definition:"

Workaround: Change the data type in the RPC declaration from "decimal" to "double".

Color changes are tied to 3d system color changes; were tied just to window text changes.

Changed 5.0.04 and 6.0 to allow NULL string variables to be passed to external dlls. The windows API calls expect a NULL C pointer and there's no other way to provide that except SetNull. ***NULL doubles and decimals, etc will still be disallowed.

OLE control's datachange and viewchange events do not get triggered. This is critical to deployment and upgrading to 5.03

The events: ViewChange, DataChange, Close, Rename, and Save are all optional events which the OLE server may or may not implement.

There are also no rules about when the OLE server should call these events when they even are implemented.

Hence, the problem is with the OLE server, not PowerBuilder.

This is general to pb but in this particular case there is a regression occuring

This issue is now fixed using pbroi050.dll from pb 5.0.04_36 or later.

From the table painter, a properties sheet can be accessed which has a Foreign Key tab page. The New button on that page is grayed out, although it can be activated by double-clicking.

Steps to recreate:

1) Enter the table painter (with a new or existing table).

2) Go to the properties sheet.

3) Go to the Foreign Key tab page.

The New button is grayed out. However, if it is double-clicked, it works.

The button should not be gray, and it should work with a single click.

Opening a dddw (one of several) on a grid, causes the

Drop Down portion of the first DDDW to display away from the drop down arrow. The items in the DDDW cannot be selected.

Workaround: Do not Modify() column color

(0001): Error C0174: Type of actual parameter cannot be ancestor of formal reference parameter type.

PER DEVELOPMENT:We won't be fixing this in 5.0--the compiler flaw that allowed this behaviour in the first place has been corrected in 6.

Solution: Logging was not turned on for the database.

Related to "Saved Data" in the dataobject since similar behavior is seen with DDDWs when you save data in the PARENT data object.

Workaround: For the DDDW column, you must join the table and column for the Display value into a second datawindow - one you use just to use Save As.. ".PSR".

DEVELOPMENT: Cannot recreate - fixed during 6.0 beta.

Workaround: Insert an empty row in the destination datawindow control first.

dw_1.dataobject="test.psr"

ds_1.dataobject="test.psr"

dw_1.insertrow(0)

dw_1.object.data = ds_1.object.data

PER DEVELOPMENT: Fixed by a different issue's code change.

32 bit Short, Win 95b ( plus) - Development pb 5.0.03

keywords:: ole automation .. office 95.. win 95b

problem description::

If you preview a dw that is OLE style (Excel Spreadsheet) you hang if you try to exit the preview.

This does not happen on my machine. Only happens with win 95B (upgrade version of win 95) and office 95. With pb 5.0.02 32 bit the DW does not hang

I use office 97 and win 95 Gold. I get no data retrieved although it states at the microtab section at the bottom of the DW that 30 out 60 rows retrieved.

There is no problem whatsoever with NT 4.0

The steps that I do to make this occur is:

1. Hit the table painter icon.

2. Select new.

3. Name the column.

4. Give the table a data type.

5. File --> Save.

6. Name it.

The error message reads:

{

Create Table Definition

Stop!

The following syntax pattern provided in the pbodb60.ini file is incorrect:

Column Element.

}

I tested the same steps on this version in Win95 and NT4 and nothing abnormal.

PER DEVELOPMENT: This specific problem dealt with importing text files with currency data. The files had been exported in US currency and were being imported with settings changed to another currency which used the comma decimal separator. The problem is that given a particular countries settings of decimal points and thousands seperators there is NO WAY we can tell or translate from a text file stored with another countries settings. This is ONLY a problem for text files, a DB wouldn't actually store the '.' or ',' in the data.

USA 119,989.27

International 11.998.927.00 instead of 119.989.27 or sometimes you cannot type anything in at all as numbers spread all over the mask.

POSSIBLE WORKAROUND: if the mask is ###,###.00, then this mask works fine with international settings, so long as this format is acceptable for the various international settings you may be using.

This is the danger of using text files for import. Having different decimal separator settings does have an effect.

/lib/dld.sl: Unresolved symbol: opinit (code) from /pwrs/pb6_162/pb6/bin/libpbo7360.sl

IOT trap (core dumped)

Short Description: When adding text to window controls, controls (check boxes,
command buttons, radio buttons, singleline edits, edit mask) do not resize appropriately.
Works ok on windows. Problem was first discovered in PB 6.0 beta.

PER DEVELOPMENT:

This is not a bug for syb, 4.0x Dblib Database servers. The primary key implementation was not 'really' supported in this version of the servers. The 4.0x servers only used the sp_primarykey stored procedures for documentation/user intent puposes. The 4.0x server did not enforce any constraints, as constraints & rules were not available in the database itself in the 4.0x version of the Server for a primary key.

If one uses MS SqlServer 6.0 (or higher) the primary key constraint is supported and enforced as expected. (ie. If one executes the steps above in SqlServer 6.0 or higher,, then one will receive the error messages: "8111 Attempting to define a primary key constraint on nullable column in table.")

This excerpt is right from the MSDN: "calling sp_primarykey and sp_foreignkey simply creates entries in the syskeys table and Microsoft SQL Server does nothing more with them." That is if one executed a sp_primarykey in a 4.21 server in ISQL (ie. No PB), the database would not enforce the primary key 'rule', where a dup column or null values, would be created by the DB.

I have included more information per the MSDN Article below

==================================================

The following is in the MSDN Article, 8/12/96, Migrating your Database from MS SqlServer 4.21 to 6.5:

Syntax to Create a Table and a Primayr & Foreign Key

============================================

CREATE TABLE Article (ArticleID PKID, ArticleType TypeCode,

AuthorID PKID, Title StdDescription NULL)

exec sp_primarykey Article, ArticleID

exec sp_foreignkey Article, Author, AuthorID

Here we have created the two tables, declared a primary key on each, and declared a foreign key in the Article table. Unfortunately, defining these keys does not mean that the database will enforce them! In version 4.21, primary and foreign key declarations are for documentation purposes only; calling sp_primarykey and sp_foreignkey simply creates entries in the syskeys table and Microsoft SQL Server does nothing more with them. To enforce the primary keys, a unique index must be created:

CREATE UNIQUE CLUSTERED INDEX XPKAuthor

ON Author (AuthorID)

Steps to recreate:

1) If employee_copy exists, drop the table.

2) Create a pipeline to copy the employee table to employee_copy.

3) Execute the pipeline.

4) Change the option to Replace.

5) Execute the pipeline.

If you open the table in the database painter, it will show up as having no columns.

Customer's Code:

int i,j

i = 0

j = 0

for i = 1 to 4

for j = 1 to 4

ole_1.object.datagrid.setdata(i,j,double(i*j),0)

next

next

WORKAROUND - Change the Code to the following:

int i,j

double d <---------

i = 0

j = 0

for i = 1 to 4

for j = 1 to 4

d = i*j <---------

ole_1.object.datagrid.setdata(i,j,d,0) <---------

next

next

12/11/97 Update from Development:

Workaround is to assign the double to a variable and use that variable in the OLE call. (see above). Development will be looking into possible means of correcting this architectural problem in a future major release, it cannot be addressed in a maintenance release.

i.e. acitvedocument.bookmarks[2].select

However the Microsoft Visual Basic for macros, allow for the reference of bookmarks through their names as well.

i.e. activedocument.bookmarks("temp").select

which the above does not work in Powerbuilder

Error: "Error calling external function bookmarks at line xx in event xxxx..."

Database=Oracle 7.3, DBMS=O73. Only happens on 16bit with O73 driver and numeric columns and DisableBind=0 (default). Inserting NULL into numeric column causes a gpf in PBO73050.DLL.

32 bit only {pb 5.0.02_11, pb 5.0.03_35, pb 5.0.04_31} panther version 186

keywords:: OCX .. mapocx from chigago Corp.

Problem description::

The OCX ( Mapocx.ocx ) is a control from Chicago Map corp. This OCX will register in the registry but PB cannot display it on a window. No OCX properties can be displayed for it. It is completely non functional within powerbuilder.

To recreate

1) after connecting to a database, go into the DB Painter and select menu options Object...New...Table.

2) Upon arriving in the table painter, your cursor is in the ColumnName field and you can immediately type the first column name - do so.

3) Tab to the DataType field and assign a datatype to the column.

4) Hit the enter key. You are now presented with a new column template for defining your second column. If you begin typing, the characters don't show up anyplace - none of the fields has focus. ColumnName should have focus - it does on Intel and in PB5_230i on Solaris.

Workaround:
Triggering the OpenSheet works fine, and using a local variable of type w_sheet as the sheetrefvar also works:
w_sheet lw_sheet
Post OpenSheet(lw_sheet, w_frame)
Or create a user event or function that triggers the opensheet(w_sheet, w_frame) and post that user event.

Per Development (11/10/97):
This is a problem with Posting with a global variable representing the window (e.g. w_sheet). By the time the posted function runs, the window class has been garbage collected and removed from memory. A work around would be to put the call to openSheet() in a function and post the execution of that function. Another alternative (but not recommended) is to turn off garbage collection until the sheet has appeared. The same problem would happen in 5.0 if a window were closed after the POST call as that is when 5.0 ran its form of garbage collection. This would require architectural changes to force the window class to stay in memory and thus needs to be deferred to a future release.

When opening a window with a listview control, the 3rd item in the listview is selected. When trying to arrow from one item to another, using the keyboard arrows, the up and down arrows do not work correctly. If an item is selected using the mouse (clicking on it), the arrows then work fine.

Create a window with a listview.

Add 5 items to the listview arranged as shown above.

Add the above code to the window open event.

When the window opens, you can see that the 3 rd item is selected. If you type the up arrow the focus does not change (If you were at the 1st item the up arrow would not move you up, there is no place to go up from item 1). If you type the down arrow, focus does not change. (If you type the down arrow for item 1 it would move you to item 3, but since item three already has focus, nothing appears to change. At this point the focus and the internal index are both at item 3 and the arrows work fine from this point on.

Development: Not a bug.

We saveas .txt files the way we (i.e. dw) would want to read them back in ( if one was to do an importfile on the file). Therefore,

let's say you have a .txt file with the following three lines:

"ABC"

""ABC""

"""ABC"""

and that this file was being imported into a dw with one column. The column value for

row 1 = ABC (we read quoted string contents and throw away the quotes)

row 2 = null (we read the first quoted string which is null and ignore the rest)

row 3 = "ABC" (we have special logic to understand that these are quotes that matter to the string)

Keywords:: modify , describe, datawindow "incorrect syntax at line XXX"

Problem description::

When you do a modify on a variable that has been associated with a describe function you get an error. "Incorrect syntax at line XXX"

Eg ii_instance = dw_1.Describe("DataWindow.Table.Select")

then do a modify

dw_1.Modify("DataWindow.Table.Select='ii_instance")

you get an "Incorrect syntax at line XXX"

This is not database specific. The modify function should be able to use the syntax from the describe.

Also using a specified sql statement for a syntaxfromsql command that is valid should be valid when applying the same statement to a modify. Instead the result is "Incorrect syntax at line XXX". This should not be the case. The modify statement should be able to recognise the syntax.

Checking out objects to 'work' pbl, after a certain number of objects an error occurs: "cCheckout Object : client rpc error out of memory" and objects if using NT3.51.

Problem only occurs with PBEOL050 and Object Cycle 2.0 on the Win 95 or NT platforms. Not a problem in Win 3.1 or in PB 6.0.

Fix requires changes to both PBEOL050.DLL and OCC020.DLL.

1.) An option to control whether or not an owner prefix is used

2.) An option to control what value for the owner prefix is applied, if used.

Operating system and version of Powerbuilder

win nt 4.0 only, 32 bit only . pb 5.0.03 and pb 5.0.04.

Problem description.::

Printing from a dw printpreview in landscape mode. When the user prints in the print preview of his datawindow 400 pages of the same data are printed out instead of 1. Also when you run the window the same thing happens .Normaly there is one page of data from the dw that should be printed. Instead there are around 400 pages being printed out of the same data.

This only happens when newspaper columns are set to 2. when they are set to 1 there is no problem.

10-21-97:: Adeering . I tested out pbdwe050.dll on the win nt 4.0. This dll is from build 65 of pb 5.0.04. I optimised the pbl and did some checks to see if pbdwe050.dll would break anything else. It did not. I will now give this dll to the customer.

The word "structure" is misspelled in the following string "Unable to find an available protocol driver strcuture".

PER DEVELOPMENT:

Confirmed that this error is coming from the server; thus, we cannot fix this misspelling: sqldbcode 20012, 'Unable to find an available protocol driver strcuture.'

This issue only occurs with 16 bit clients.

Problem fixed in 5.0.04, but still a problem in 6.0.01

Getting error message, "System error sharing violation on drive <X>:". PowerBuilder is installed locally but the pbl's are on the Novell network. If the pbl's are brought local or an NT server is used this error does not occur.

1. Hold left mouse down.
2. Press right mouse and pick "paste" from menu.
3. Let up on right mouse
4. Let up on left mouse. GPF.

This does not happen in 6.0.00 and beyond.

MsQuery 32 bit gives the error "syntax error in from clause". Seems like a driver issue.

Error: The Property 'entryid' was found in class 'mailrecipient', but insufficient rights are available to access it.

Solution:

Documentation bug in Online help:

ODBC DECLARE and EXECUTE

DECLARE logical_procedure_name PROCEDURE FOR

procedure_name

({@param1 = value, @param2 = value2, ...})

{USING transaction_object};

---> For ODBC parenthesis should not be used -

In general, when this retrieval argument issue was investigated in other drivers, it was mentioned that users should use "staticbind=0" to re-describe the result set - effectively forcing the developer back to 3.0/4.0 behavior to get rid of binding problems.

NOTE: When DBPARM is left blank (meaning staticbind should default to "1") we can still see in the trace that a DESCRIBE is done on the result set and yet we still get the error "Column lists do not match". Staticbind=1 should be the default.

Example: Datawindow "a" is the parent and shares with "b" and "c"

In the rowfocuschanged of "b", there is the function call

"c".modifiedcount(). This causes a GPF in PBDWE050.DLL 0137:11589d1d when a row is deleted from the parent or "a" datawindow.

Workaround: Easy. Put the Modifiedcount() in a posted event from the rowfocuschanged event in datawindow "b"

OS and version: win nt 4.0 , win 95 32 bit and 16 bit {pb 5.0.03}

Keywords:: External function call

Problem description: Customer is trying to make a call to an external function she wrote.

Customer gets the following error

"Unsupported argument type in function ml_net_addconnection in herdll.dll". This error message did not appear in pb 5.0.02.

This issue was fixed with pbrte050.dll w.r.t pb5.0.04. I also tested this out in pb 6.0 build 235.

A -1 error is returned, but all the custoemr wants is to be able to use this function call ml_net_addconnection

This works in 5.0.04_76 & 6.0

/*

//*

*/

This is a problem for the customer because of Orca. The Orca import //*, when enclosed in a comment block, causes the compile entry import function to hang.

Errors:

C0035 Unclosed Comment and C0031 Syntax Error.

Workaround: Go through the file and globally replace any '//*' with, '// *'.

Error: "Error accessing external property selected at line 'x' in 'y' event of object 'z' "

Workaround:

Since the dot notation defaults to the Primary buffer, the workaround is to omit the keyword:

i.e.

DO NOT use dw_1.object.columnname.primary.selected, but

dw_1.object.columnname.selected

or

dw_1.object.columnname.current.selected

PER DEVELOPMENT:

This is not a bug.

The limitation is established by a configurable registry parameter which defaults to 17.

The Registry value is as follows:

HKEY_LOCAL_MACHINE

\SYSTEM\CurrentControlSet\Services\Lanmanworkstation\parametersMaxThreads

Add this parameter in the registry and set = to desired limitation.

Tested this on 2 client machines and it fixed it on both.

PER DEVELOPMENT: No plans to address this issue. Future releases may make this obsolete.

Allow AutoSize on bands other than detail. Often it is desirable to use the detail band only for summarizing data and it is never displayed. The actual data is displayed in group footers with totals. If the data is a large variable text, there is no way to adjust the height of that band.

em_1.text = "0"

the modified event doesn't fire until you put focus on an editmask then force a losefocus by way of clicking or tabbing.

The documentation states:

Modified event doc:

When a control has been changed AND loses focus (becomes inactive).

This makes it harder to change data in a window editmask programmatically since the modified event will fire sometimes when you don't expect.

16 bit only {pb 5.0, 5.0.02_11,pb 5.0.03_35, pb 5.0.03_41}

Keywords:: Rich text edit control,, bad display text format, copyRTF

Problem description::

This is win 3.11 only. both 16 and 32 bit PB on win 95 or win nt 4.0 have no problems.

Customer is using a RTE control. He retrieves a column from the database that has all of the rich text formatting codes. A single quote or ** appear in the display of the RTE . If you check the text right after retrieving, there is no single quote or **.

It is only when you issue a copyRTF that the problem occurs.

Happens in development and in run time also.

It appears that PB 5 calculates the 1/1000th inch differently than version 4. Creating a dw on version 4 and migrating to 5, the paper size goes from 8.5" wide to only 8". This is causing pages to wrap.

There seems to be a difference in PBRTE050.DLL in the deployment kit vs. \sys32 or \sys (dev) directory. The dll in \pb5i32\dk in some cases will cause a page fault in PBSHR050.DLL@0173:1111198e.

PER DEVELOPMENT:

New deployment kits available on ftp site.

Files: 503dk16.zip and 503dk32.zip.

Moving the code out of the pbm_dwngraphcreate event gets rid of the GPF.

6.0.00 and 6.0.01 work fine.

Customer has coded some print functions on the Server app. When called from the Server, these print functions work fine. However, when these Server print functions are called from the Client, they fail.

Using "ShowHelp ("help file",Topic!,topicnum)" fired from command button. Works in development and p-code but not in machine code executable...Error displayed:

"Topic does not exist, contact your application vendor for an updated help file (129)"

(This worked in PB 4.0.0x)

Workaround: Retrieve data first using Retrieve(), then call CrossTabDialog() function.

Back in version 5 you were able to run your app with the Debugger and utilize Ctrl-S to step through each line. Now you have to use the mouse to click on a toolbar icon. I know it can't be that hard to bring back this nice time-saving feature.

PER DEVELOPMENT: If you try calling this from a C program you get the same results. This is not a PowerBuilder bug.

Workaround:

Create a separate transaction object with AutoCommit=OFF and use this one to handle transaction processing.

User has been set up as with the Role of a user and can clear the registration of an object even though a message is given stating that 'User does not have permission to invoke function cDeleteObject'

Developer: Already fixed earlier in 6.0

971104 SLIE - Not reproducible with US settings. Can use Norwegian, Swedish or German settings to reproduce, where comma is used as decimal separator.

- does not occur under 5.0.04_71, 6.0.01_339

In 16 bit, the GPF is in PBSYC050.DLL @ 0003:39ce (5.0.03) and in 0003:3a9e (5.0.04.70) and PBVM60W.DLL @ 001B:17ae in 6.0.01 build 314.

Workaround: Use Postevent from the DBerror to do the disconnect.

32 bit only {pb 5.0, 5.0.02_11, 5.0.03_35, 5.0.04_41}

Keywords:: NVO , bad resource performance, create,destroy

Problem description::

Customer creates and destroys a matrix of NVO's and the performance is effected as a result. I tried this out using 2 NVO's. I found the following result as the customer and the TSE

# of nvuo CREATE (sec) DESTROY (sec)

1 1 1

2 3 4

You cannot run this app in 16 bit mode on any platform as you will result in a gpf in pbshr050.dll.

Checked with resource monitor on NT 4.0 , and there is no major memory loss. All memory is freed up when you stop the application.

In powerbuilder 4.0.07 there is no memory loss.

SOLUTION:

This is as designed in all PB versions 5.0 and later. When you declare a static array we must initialize all elements of the array. In 4.0 the initialization would be done the first time you access each element of the array. Therefore in 4.0 you pay the price throughout your app where in 5.0 you pay for it up front. Use dynamic arrays to get around this.

1997/////-...8///-13

Possible workarounds:
Don't create the profile with username/password. Users must log in each time. You can set the pb.ini to read-only but this causes other problems (better with InfoMaker).

are lost. Copying and pasting in and out of RTF, fonts and styles are also lost or do not appear.

This issue has beentested in both win 95 and win nt 4.0 . Pb 5.0.04 build 54 fixes this problem.

Note when running the debugger you will see the dataobject been assigned to the datastore but will result in a gpf in pbrte050.dll. If you apply the above dlls the problem will not occur.

dw_1.object.data=dw_2.object.data

Development: Fixed around 6.0.01 build 360 with other debugger fixes.

Footer band of composite report overlays the trailer band of a child report if it's at the bottom of a page.

Can we PLEASE have a property to control the background colour of 3D checkbox and radiobutton edit styles in datawindows so we can provide the user with the proper visual cues when columns of this type are readonly or disabled. (see checkbox and radiobutton window controls.) If the datawindow is the preferred interface control then we should have at least the same functionality as the window control provides by version 6.

This behavior would have to include the protect = 1 or taborder = 0 attribute effectively showing the "disabled" look as well.

CrosstabSum(1, 2, "@year")

the customer is using several columns but it shouldn't GPF

This is not a bug..

With any newly added columns to a tabular or grid datawindow, you should place a static text above the column in the header band. The problem is: this static text representing the header has NO "name" attribute and that's what makes the sorting on headers work.

When a datawindow is newly created, any columns automatically get headers (grid, tabular) with the name <columnname>_t as their "name" attribute.

Ex:

1. "dept_head_id" column is added after the datawindow was originally created.

2. A static text placed above "dept_head_id" was typed in as "Department Head ID".

This static text has no "name" attribute - it's NULL - and that's what the of_setcolumnnamesource(2) uses. The naming convention would be:

<columnname>_t

Name attribute: dept_head_id_t << name of the static text in header associated with dept_head_id

Text attribute: "Department Head ID" or anything you want

ex: dw_1.print(true)

Workaround: Use dw_1.print(true). You must explicity set the flag to get the cancel dialog. OK on other platforms.

GPF in pblib050.dll when you try to import the exported file.

5.0.02 - pblib050.dll @ 0000058c

5.0.03 - pblib050.dll @ 00000416

5.0.04_48 pblib050.dll @ 0000058c

preserved, PB always reverts foreign key name to '[foreign key 1]'. This system generated name is invalid and subsequently causes errors

when user attempts to modify foreign key definition.

THIS IS AFFECTING TEST AUTOMATION SCRIPTS...

PER DEVELOPMENT: This is not a bug. Syb 4.0x did not support foreign key (or primary keys for that matter) assigned key names. Please bug number 44811's description for more information on syb 4.0x's support of foreign and primary keys. (This information is also available in the MSDN.)

This behaviour is the same in 4.0x, 5.0x and 6.0x. Since the database didn't provide/save a foreign key name, the PBSYB*.dll driver was coded (and only this driver, I double checked), to give a default foreign key name of '[foreign key 1]'. The driver was also coded to give a default primary key name of 'primary'. This is true only for this dialog and only for the SYB driver 4.0x.

NOTE: The display of the new report is fine. It's the printing that's truncating the data - so you must printout the psr.

Computed fields are moved. Numbers are cut off as are characters.

Using the report painter in Powerbuilder and the "Save Rows As..." Powersoft Report using a PSR EXACTLY duplicates the PSR for viewing and printing.

every other windows event does.

Workaround: Use message.longparm/wordparm.

ex:

dw_1.of_Setfilter(TRUE)

ii_rc = dw_1.inv_filter.of_SetColumnNameSource(2)

dw_1.inv_filter.of_setstyle(1) // Can't use "0" for default dialog, this won't work.

dw_1.inv_filter.of_setdefaultheadersuffix("_t") // need this too.

dw_1.inv_filter.of_SetFilter(ls_null)

dw_1.inv_filter.of_filter()

of_setdefaultheadersuffix() function call must be used to tag what it should look for as a header. If you use of_setdefaultheadersuffix() and you've change what the header is called. For instance:

'dept_id' will have a default header name text object called "dept_id_t"

If you modify any of the text field "names" in the header on the window (or it doesn't find one for the corresponding column), it won't work. If you keep all the defaults, in the example shown above "dept_id", if you use the extended filter dialog, the column in the "columns" tab will be referred to as "Department ID" - ie what is called the "label" in the extended attributes or the "text" attribute of the header "dept_id_t".

Basically, internally a mapping table is built that carries all the information so of_setcolumnnamesource() works correctly.

Datawindow column database column header

=========================================================

dept_id department.dept_id Department ID

PER DEVELOPMENT: NOT A BUG: DOCUMENTATION ERROR.

Workaround:

==========

Instead of an embedded SQL SELECT statement after the UPDATEBLOB, perform a SELECTBLOB instead and use the string() function to convert the results

WORKAROUND: Whenever font properties are modified the user has to resize the object to be the size they want.

This seems to be a development environment issue only. This dialog is available by using Right-Mouse on a blank area of the datawindow in design mode and choosing Properties/Print Specifications and selecting the "size" field in the "Paper" groupbox.

When Exporting object from the library painter, the "Export Library Entry" MODAL NON-RESIZEABLE response window height is based on the size of the PB application window. You can shrink the PB application so that the response window becomes unusable since its height it too small to see all the controls on the window. The only work around is to return to the library painter, resize the PB application so the response window height will resize so you can see all of the dialog box.

Solution: Added support for special datatype support from the pbodbxx.ini when we export syntax.

The Browse button on the Database Preferences window crashes PB.

1) Enter the database painter

2) Open the Database Preferences window (Design>Options)

3) Click Browse

You get the following error:

Can't access this folder.

Path is too long.

Then PB crashes or hangs.

The uploaded test case has buttons to click which demonstrate the behavior. Each button has the same basic script except for how the initialization is done.

//////////

int li_Index, li_Count

string ls_Out

int li_Elements[]

// division and modulo fail

// addition, subtraction, multiplication succeed

li_Elements = {1 + 2}

li_Count = UpperBound (li_Elements)

for li_Index = 1 to li_Count

ls_Out += String (li_Elements [li_Index])

if (li_Index < li_Count - 1) then

ls_Out += ", "

end if

next

mle_1.Text = ls_Out

/////////

Composite dw with one report object. Criteria property for the report object does not get recognized, if one is specified in the dw painter. It will retrieve all rows in the table.

Solution: Not an issue in 6.0 since there is no 16 bit dev env anymore. This will not be addressed in 5.0.

This is NOT bug 45421. It's just not calculating the sorts at the correct time.

Workaround: Invoke either Sort() or Filter() and possibly GroupCalc() at runtime (or in painter) to repaint the computed fields. Use whichever one is less destructive.

Per Development:
This is not a bug. This is a case of an aggregate-of-an-aggregate, which has always been 'unsupported' in DataWindows.

The group 1 computed field is the sum of a computed field in the detail band.
The computed field in the detail band is a calculation using another computed field in the detail band.
The other computed field in the detail band calls the function CumulativeSum on a third computed field in the detail band.
This boils down to a Sum of a CumulativeSum -- an aggregate of an aggregate.
The DataWindow Engine does not guarantee the computation order of computed fields, and therefore aggregates of aggregates is still unsupported.

However, I did find a work-around for the test case. The work-around might work for the customer and other aggregate issues.
I built a test window for the issue. In the Open Event I entered:

dw.SetRedraw (false)
dw.SetTransObject (SQLCA)
dw.Retrieve ()
PostEvent ("group_calc")

I created a user-defined event called "group_calc", in the window. I gave it the script:

dw.GroupCalc ()
dw.SetRedraw (true)

When I ran the application, the totals displayed correctly because of the double calculations.

I turn Redraw off to avoid a "display blink".
I post the group_calc event to allow the DW to complete its setup.
The group_calc event calls GroupCalc to recalculate group aggregates.
This works like a second-pass over the data.
I then turn Redraw back on and the computed fields display correctly.

Cannot recreate in 5.0.04.55 or 6.0. Possibly related to bug 43735.

This problem does not occur in builds 247 (6.0) or 301 (6.1).

Only happens if both nonsecure and secure versions are present on the same window. Does not happen if you rightmouse click and cut. Does not happen if only one control is instantiated and deleted.

pb050.exe /p datapipeline /l c:\cir\print_50.pbl /N

PB60.EXE /P datapipeline c:\cir\print_50.pbl /o test

Workaround: Use Library Painter

entry -. properties (Right mouse button) does allow entry or modification of application comments.

Developer: The itemerror event is telling DWE to accept whatever value was entered in the column. In a numeric column if you enter and force acceptance of the character "N" it may convert to a large number. Likewise, if you force acceptance of a null string it will convert so a number depending on the system conversion routines. Any assumptions DWE makes about what to leave in the column is likely to be wrong for someone.

The Cancel option, when selected on the PVCS registration dialog, should return SCC_I_OPERATIONCANCELLED from the SCCAdd call.

The archive bit (the drum icon) is displayed for all objects (including the canceled objects).

Solution Per Development:

PVCS_SCC.DLL returns SCC_OK even though the user clicks CANCEL in their dialog.

This will be fixed in PB 5.0.04.

=====================================================

Pre-Released fix for Powerbuilder 5.0.03 or greater

Library Painter and Source Management for PVCS

=====================================================

Date: January 10, 1998

=====================================================

The zip file 'pblib503.zip' contains two maintenance DLLs for Powerbuilder Enterprise 5.0.03 for Windows 95 and Windows NT. They are compatible with the 5.0.03 and 5.0.04 maintenance releases.

To install, rename your existing source control DLLs PBSCC050.DLL and PBLIB050.DLL to *.old. Then copy the new PBLIB050.503 and PBSCC050.503 to the directory containing the development environment DLLs and rename them to *.DLL.

These DLLs contain fixes for the bugs listed below. Some are related to Powerbuilder library painter general bugs, others are specific to the Intersolv PVCS interface PBSCC050.DLL.

URL for download:

ftp://ftp.powersoft.com/pub/private/pbpatch/pvcs/pblib503.zip

===============================

NOTE: These fixes are already integrated into the Powerbuilder 5.0.04 maintenance release and in Powerbuilder 6.0.

INTERSOLV has supplied a URL to download their fixes from the Intersolv web site. Customers can obtain a copy of the "INTERSOLV PVCS Version Manager Source Code Control (SCC) Interface Version 5.1" at:

ftp://ftp.intersolv.com/download/pvcs/vm/if_32_51.zip.

=====================================================

12/15/97 TSE test results update:

Intersolv Issue: Noticed that the PVCS SCC dialogs have problems when run using Large Fonts.

The current default behavior(s):

* numeric EditMask - insert mode
* String-EditMask - overtype mode
* editmask control without a special Mask - insert mode

Use of SetMask(StringMask!, "mask" ) switch an editmask from insert to overtype mode.

dw_1.describe("button_1.type") returns "!"

dw_1.describe("button_1.text") returns the correct button text.

NOTE: Global functions in computed fields generally will slow down painting of the datawindow due to the amount of activity in the global function.

Workaround: Must modify the SQL and add the order by BEFORE the retrieve()

This issue has particular urgency for DB2/MVS users as the only way one can get the SQL code back from a DB2/MVS native stored procedure is to send it to the client in an output parm. Also, no output information can be sent back to the client (again unless one uses RPC's).

immediately after one column has been changed.

Modifying properties on datawindow columns with EditMasks causes a page fault in PBDWE050.DLL@0137:115e0906 (for 32bit), @000b:00001865 (for 16bit). This happens using using Modify() or dot.notation.

Developer: Already fixed in 6.0

As tested, it works with GetItemNumber, constants, and variables. However, when adding the numbers, it inputs the first dot notation-referenced variable, then doubles it in each dot reference thereafter, ignoring the value.

decimal counter

// all values in column_1-n are "1"

counter += dw_1.object.column_1[1]

counter += dw_1.object.column_2[1]

counter += dw_1.object.column_3[1]

counter += dw_1.object.column_4[1]

Instead of counter=3, counter is "8"

Using other variable types for "counter" makes it worse:

Type Value

long - 0

real - -5,41817766932997E+23

double - -7,74025381332799E+22

CONSTANT integer ci_value = -1

long l_local

l_local = ci_value

if l_local = ci_value then

MessageBox( "Result" , True )

else

MessageBox( "Result" , False )

end if

will always return False. However the values are the same ...

GetObjectAtPointer gives a wrong object name in some circumstanses.

The error message you will receive:

"The name @@objname is not a valid local variable name".

Solution:

The following patches below contains the fixes to the stored procedures. Development mentions that you should upgrade to Adaptive Server 11.5 before running the scripts. If you're not running Adaptive Server, you don't need to do anything.

In order to convert the existing stored procedures used by Powerbuilder 5.0 and 6.0, users should execute the following scripts in ISQL to re-write the stored procedures Powerbuilder uses.

For Powerbuilder 5.0 users

ftp://ftp.powersoft.com/pub/private/pbpatch/adaptive_server_11.5/pbsyc50.sql

For Powerbuilder 6.0 users

ftp://ftp.powersoft.com/pub/private/pbpatch/adaptive_server_11.5/pbsyc60.sql

If you force a change using the property sheet, the text in the computed field will move so it "looks" right-justified but only until you close the painter. When you re-open the painter, the problem will exist again.

NOTE: The preview is fine, it's just a painter issue with startup.

To Reproduce:

1. Create a new window.

2. Add an OLE object, word97.

3. Create a button with the following script: ole_1.linkto("c:\some word file")

4. Run the window.

5. Click on the button.

6. When the linkto is complete close the window.

Pb will gpf. It doesn't matter what OLE object you use or the manner in which you close the window (script, alt-F4, etc...).

PER DEVELOPMENT:

Each recursive call adds to the stack. Since the stack is finite and cannot be changed, you will always eventually run out of space. This limit may change from release to release and there is not much we can do about it. We will keep this in mind in the future releases to see if we can increase the limit.

6.0 build 250 GPF is in PBVM60W.DLL at 005B:15D6.

Oracle (O73)

-----------------

Column Type Save As Syntax

Date datetime

Varchar Char

Varchar2 Char

Number (5,0) Decimal (16,0)

SQL Anywhere

---------------------

Column type Save As syntax

varchar15 char(15)

numeric decimal(16, 0)

integer float

smallint float

double float

long varchar char(32767),

tinyint float

System11(SYC) - Use alldatatypes on tsdata.

---------------------

Column Type Save As Syntax

varchar(10) Char(10)

int float

smallint float

tinyint float

real float

text char(32767)

bit float

money decimal (16,4)

smallmoney decimal(16,4)

smalldatetime datetime

Informix 7.x using IN7

-----------------

Column Type Save As syntax

decimal16_0 decimal(16, 0)

integer float

smallint float

real float

money16_0 decimal(16, 0)

varchar10 char(10)

serial float

interval char(0)

Microsoft SQL Server 6.5

----------------------------------

Column Type Save As syntax

varchar10 char(10),

int float,

smallint float,

tinyint float,

real float

text char(32767)

bit float

decimal10_2 decimal(16, 2)

numeric10_2 decimal(16, 2)

money decimal(16, 4)

smallmoney decimal(16, 4)

smalldatetime datetime

identity integer float

Restart not functioning correctly. It states in the on line help that restart is normaly called from the idle event of the application object. Using it here closes down db activity but will not retart the app as it is suppoesed to do. Calling it from any other event will not close down db activity and restart .

Definition is "Stops the execution of all scripts, closes all windows (without executing the scripts for the Close events), commits and disconnects from the database, restarts the application, and executes the application-level script for the Open event"

val = string ( theRecs.Fields.item(theRecs.tabledesc.fieldname[i]).Value) )

"theRecs" is an OleObject associated with MapObjects. The user is getting a value

out of a Dbase table to place that value into a string and into the Grid Control.

It GPFs whenever the value out of the table was a long or floating point number.

Workaround: Break up the statements.

long lnum

string gridText

lnum = theRecs.Fields.item(val).Value

gridText = string(lnum)

For the MapObjects example:

mapLayer.GeoDataset = dataConnect.FindGeoDataset(fileName)

"mapLayer" is an OleObject associated with MapObjects, as is "dataConnect". This is

what I had to do to get rid of the GPF:

OleObject geoData

geoData = CREATE OleObject

geoData = dataConnect.FindGeoDataset(fileName)

mapLayer.GeoDataset = geoData

.

.

.

DESTROY geoData

Development: 4.0 had the behavior of correctly putting up a "datawindow error" messagebox when the dw_1.Update() returned a -1 - when the user coded the SQLPreview with a "setactioncode(1)" (stop processing).

5.0 had the regression where update() no longer returned a -1 with "return 1" in the SQLPreview. This has been fixed in 5.0.04. Users should remember to code a "return 1" in the DBerror event to supress the "datawindow error" messagebox when cancelling an Update() with a 'return 1' in the SQLPreview.

Cannot recreate using 5.0.04 and 6.0 Beta.

Run datawindow with nested report in dw painter or runtime will cause a page fault in PBDWE050@0137:115f3b8c (in 32bit) @004:00004bb8 (in 16bit). This happens under the following: Nested report has a value in Criteria Property AND Detail Band Autosize Height is checked on NESTED REPORT. Take one or the other off, and it will not cause an error. This also happened in 5.0.02. Works in 5.0.03 and 5.0.04 through build #48.

The name of the column is not defined

WORKAROUND: Easy - delete key and recreate as clustered.

CANNOT RECREATE using build 230.

The following code will GPF (PBVM60.DLL@0137:00688a74) at runtime if one of the datawindow columns is of type 'decimal' and one of the values in that column is a NULL: struct = dw_1.object.data.[x1,y1,x2,y2]

NOTE: This problem does not seem to happen if other datatypes are used and have NULL values.

Fay: Verified it works correctly in 6.0 Build 247

To Reproduce:

1. Create a window with an ole object (Excel '97 worksheet) and a button.

2. Add the following script to the button:

int rc

ole_1.Object.Application.Range("A1:E5").Value = 100

rc = ole_2.linkto("add your file name here.xls")

parent.title = string(rc)

3. Run the window.

PB crashes when it gets to LinkTo. I have tried this with Word '97and there was no problem. It seems to only occur with Excel '97

activating Excel OLE objects.

Using Win '95 and Office '95: Offsite activation is OK, Inplace is broken.

Using Win '95 and Office '97: Offsite activation is OK, Inplace is broken.

Using NT4.0 and Office '97: Both Offsite and Inplace are broken.

To Reproduce:

1. Create a window with and OLE object (Excel file) and two buttons.

2. Label one button Offsite and put the following script under it:

ole_1.activate(offsite!)

3. Label the other button Inplace and put the following script under it:

ole_1.activate(inplace!)

Run the window and click on the buttons. You will notice the strange behavior.

Parts of the spreadsheet will be missing when you try inplace activation

(headings for the rows and columns). When you try offsite activation Excel will

come up but without the spreadsheet. The behavior is rather unpredictable.

PER DEVELOPMENT: Applying 11.1 server appears to clear up this problem.

3/25/98 FBASHAW: Verified fix in Build 385.

create procdure proc_role2(in @role_type as /* (Using Ms SqlServer) */

Select count(*) from employeechar(10))

or

create function DBO.proc_role2(in @role_type char(10))

returns integer /* Using Sql Anywhere */

on exception resume

begin

return(0)

end;

The create procedure works correctly, but when I go to the DBA Notepad and review my syntax using the Design/Procedure window, one letter at the end of every line is truncated. Why?

PER DEVELOPMENT: Cannot Recreate in 5.0.04, build65 or 6.0.00.239.

After highlihted "mm-yy" characters in the edit-mask and using the editmask.cut () function, you cannot insert more than 2 characters after the "first cutted position".

// ds is a datastore

str = ds.object.data

actually CORRUPTS the datawindow buffer (on the right side of the assignment) for the decimal values. Doing a GetItemDecimal against the datastore in the decimal field (if it contains a number like 0.22222 - GetitemDecimal returns 28131743735386124.)

Solution: There was a somewhat related structure bug with preserving nulls fixed around build 30 in PBRTE050.DLL. Using 5.0.03 with 5.0.04.61 PBRTE050.DLL fixes this problem in the runtime environment.

Verified that the application does work correctly in pb 5.0 and pb 5.0.02.

Solution:

Developer: The PowerScript syntax was incorrect and earlier versions of PB did not care.

Specifically an example like:

g_ole_f1book.ODBCConnect(pConnect, true, pRetCode) // wrong