perl-win32-gui-hackers Mailing List for Win32::GUI

Jez,

Thanks for the report and taking the time to look into this.  I've not
had a chance to look at what you've done yet, but should get some time
in the next couple of weeks.

Thanks,
Rob.

2008/4/13 Jeremy White <jez...@ho...>:
>
>  Hi,
>
>  I've created the following tracker item for this issue:
>
>  http://sourceforge.net/tracker/index.php?func=detail&aid=1941264&group_id=16572&atid=116572
>
>  I think the solution is straightforward but don't want to commit the code yet.
>
>  As a side, I use the Win32::GUI::Coolbar module from Robs homepage:
>
>  http://www.robmay.me.uk/win32gui
>
>  and it needs updating to run under 1.6 (constants). I don't mind updating this module and would support it's move into the core.
>
>
>  Cheers,
>
>  jez.
>
>  > From: jez...@ho...
>  > To: per...@li...
>  > Date: Sun, 6 Apr 2008 11:21:53 +0000
>  > Subject: [perl-win32-gui-hackers] Crash with perl 5.10 and Win32::GUI 1.6
>
>
> >
>  >
>  > All,
>  >
>  > I am getting a crash when using Perl 5.10 and Win32::GUI 1.6. The same application doesn't crash under 5.8.7 or 5.8.8. I've tracked down the "problem", but I'm not convinced it's the actual underlying issue. The crash was occurring at:
>  >
>  > GUI_Helpers.cpp line 757 proc WindowsHookMsgProc
>  >
>  > When I was selecting from the main menu of the application. The problem is the line:
>  >
>  > arrayref = av_fetch(perlud->avHooks, WM_TRACKPOPUP_MSGHOOK, 0);
>  >
>  > where:
>  >
>  > perlud->avHooks
>  >
>  > contains zero (NULL)
>  >
>  > Simply adding:
>  >
>  > if (perlud->avHooks !=NULL ) {
>  >
>  > before the call to the av_fetch stops the crash and the application runs normally.
>  >
>  > A potentially related issue is that I'm also getting:
>  >
>  > Can't call method "STORE" on an undefined value during global destruction.
>  >
>  > When the application exits. Again, this doesn't happen under 5.8.7 or 5.8.8. I am therefore suspecting a reference count issue somewhere? Any thoughts? I am going to try and create a simple test case.
>  >
>  > Cheers,
>  >
>  > jeremy.
>  >
>  >
>  >
>  >
>  > _________________________________________________________________
>  > Amazing prizes every hour with Live Search Big Snap
>  > http://www.bigsnapsearch.com
>  > -------------------------------------------------------------------------
>  > This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
>  > Register now and save $200. Hurry, offer ends at 11:59 p.m.,
>  > Monday, April 7! Use priority code J8TLD2.
>  > http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
>  > _______________________________________________
>  > Perl-Win32-GUI-Hackers mailing list
>  > Per...@li...
>  > https://lists.sourceforge.net/lists/listinfo/perl-win32-gui-hackers
>  > http://perl-win32-gui.sourceforge.net/
>
>  _________________________________________________________________
>  The next generation of Windows Live is here
>  http://www.windowslive.co.uk/get-live
>
> -------------------------------------------------------------------------
>  This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
>  Don't miss this year's exciting event. There's still time to save $100.
>  Use priority code J8TL2D2.
>
>
> http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
>  _______________________________________________
>  Perl-Win32-GUI-Hackers mailing list
>  Per...@li...
>  https://lists.sourceforge.net/lists/listinfo/perl-win32-gui-hackers
>  http://perl-win32-gui.sourceforge.net/
>

Hi,

I've created the following tracker item for this issue:

http://sourceforge.net/tracker/index.php?func=detail&aid=1941264&group_id=16572&atid=116572

I think the solution is straightforward but don't want to commit the code yet.

As a side, I use the Win32::GUI::Coolbar module from Robs homepage:

http://www.robmay.me.uk/win32gui

and it needs updating to run under 1.6 (constants). I don't mind updating this module and would support it's move into the core.

Cheers,

jez.

> From: jez...@ho...
> To: per...@li...
> Date: Sun, 6 Apr 2008 11:21:53 +0000
> Subject: [perl-win32-gui-hackers] Crash with perl 5.10 and Win32::GUI 1.6
>
>
> All,
>
> I am getting a crash when using Perl 5.10 and Win32::GUI 1.6. The same application doesn't crash under 5.8.7 or 5.8.8. I've tracked down the "problem", but I'm not convinced it's the actual underlying issue. The crash was occurring at:
>
> GUI_Helpers.cpp line 757 proc WindowsHookMsgProc
>
> When I was selecting from the main menu of the application. The problem is the line:
>
> arrayref = av_fetch(perlud->avHooks, WM_TRACKPOPUP_MSGHOOK, 0);
>
> where:
>
> perlud->avHooks
>
> contains zero (NULL)
>
> Simply adding:
>
> if (perlud->avHooks !=NULL ) {
>
> before the call to the av_fetch stops the crash and the application runs normally.
>
> A potentially related issue is that I'm also getting:
>
> Can't call method "STORE" on an undefined value during global destruction.
>
> When the application exits. Again, this doesn't happen under 5.8.7 or 5.8.8. I am therefore suspecting a reference count issue somewhere? Any thoughts? I am going to try and create a simple test case.
>
> Cheers,
>
> jeremy.
>
>
>
>
> _________________________________________________________________
> Amazing prizes every hour with Live Search Big Snap
> http://www.bigsnapsearch.com
> -------------------------------------------------------------------------
> This SF.net email is sponsored by the 2008 JavaOne(SM) Conference
> Register now and save $200. Hurry, offer ends at 11:59 p.m.,
> Monday, April 7! Use priority code J8TLD2.
> http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
> _______________________________________________
> Perl-Win32-GUI-Hackers mailing list
> Per...@li...
> https://lists.sourceforge.net/lists/listinfo/perl-win32-gui-hackers
> http://perl-win32-gui.sourceforge.net/

_________________________________________________________________
The next generation of Windows Live is here
http://www.windowslive.co.uk/get-live

Hi,

I've spent some more time on this. I've created a bug report on sourceforge so this isn't lost:

http://sourceforge.net/tracker/index.php?func=detail&aid=1941241&group_id=16572&atid=116572

Ok - it's not a hook issue, it looks like its just the normal destruction process which seems different under 5.10. Any changes to the code below (such as commenting out a line) changes the destruction order and thus the message isn't reproduced. Spent a while looking through the code, (Win32::GUI::WindowProps) and I admit I'm stumped:)  Is all this functionality to support  accessing the internals of an object? I thought that was deprecated functionality?

As a side, I couldn't get Win32::GUI to crash because of this issue but it does mean that the script doesn't finish correctly - which in turn could have other issues (such as with PerlApp).

Cheers,

jez.


> From: jez...@ho...
> To: per...@li...
> Date: Sun, 6 Apr 2008 17:54:27 +0000
> Subject: Re: [perl-win32-gui-hackers] Crash with perl 5.10 and Win32::GUI 1.6
> 
> 
> Ok: 
> 
> Found a simple way to reproduce the error:
> 
>   Can't call method "STORE" on an undefined value during global destruction.
> 
> It doesn't seem related to the crash. The stripped down example below is odd as most of the code doesn't make sense (!) but if you comment out any of the lines it doesn't show the error. The error happens with the stock Win32::GUI 1.6 ppm and with the mingw built source under Perl 5.10 (activestate build 1002). I suspect the hook method - but I'm lost...Thoughts?
> 
> Cheers,
> 
> jeremy.
> 
> ########################
> use Win32::GUI();
> use strict;
> 
> my @rows;
> for (0..250) { push @rows,'abc'; }
> 
> my $win = new Win32::GUI::Window (-name => "win");
> $win->Hook(32768+3,sub {});
> my $label   = $win->AddLabel(-name => "abc");  
> 
> #print "$win  $label\n";
> sub Message3 {$win->Hide();};
> 
> _________________________________________________________________
> Get Hotmail on your mobile. Text MSN to 63463 now!
> http://mobile.uk.msn.com/pc/mail.aspx
> -------------------------------------------------------------------------
> This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
> Register now and save $200. Hurry, offer ends at 11:59 p.m., 
> Monday, April 7! Use priority code J8TLD2. 
> http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
> _______________________________________________
> Perl-Win32-GUI-Hackers mailing list
> Per...@li...
> https://lists.sourceforge.net/lists/listinfo/perl-win32-gui-hackers
> http://perl-win32-gui.sourceforge.net/

_________________________________________________________________
The next generation of Windows Live is here
http://www.windowslive.co.uk/get-live

Ok: 

Found a simple way to reproduce the error:

  Can't call method "STORE" on an undefined value during global destruction.

It doesn't seem related to the crash. The stripped down example below is odd as most of the code doesn't make sense (!) but if you comment out any of the lines it doesn't show the error. The error happens with the stock Win32::GUI 1.6 ppm and with the mingw built source under Perl 5.10 (activestate build 1002). I suspect the hook method - but I'm lost...Thoughts?

Cheers,

jeremy.

########################
use Win32::GUI();
use strict;

my @rows;
for (0..250) { push @rows,'abc'; }

my $win = new Win32::GUI::Window (-name => "win");
$win->Hook(32768+3,sub {});
my $label   = $win->AddLabel(-name => "abc");  

#print "$win  $label\n";
sub Message3 {$win->Hide();};

_________________________________________________________________
Get Hotmail on your mobile. Text MSN to 63463 now!
http://mobile.uk.msn.com/pc/mail.aspx

All,

I am getting a crash when using Perl 5.10 and Win32::GUI 1.6. The same application doesn't crash under 5.8.7 or 5.8.8. I've tracked down the "problem", but I'm not convinced it's the actual underlying issue. The crash was occurring at:

GUI_Helpers.cpp line 757 proc WindowsHookMsgProc 

When I was selecting from the main menu of the application. The problem is the line:

arrayref = av_fetch(perlud->avHooks, WM_TRACKPOPUP_MSGHOOK, 0);

where:

perlud->avHooks

contains zero (NULL)

Simply adding:

if (perlud->avHooks !=NULL ) {

before the call to the av_fetch stops the crash and the application runs normally.

A potentially related issue is that I'm also getting:

Can't call method "STORE" on an undefined value during global destruction.

When the application exits. Again, this doesn't happen under 5.8.7 or 5.8.8. I am therefore suspecting a reference count issue somewhere? Any thoughts? I am going to try and create a simple test case.

Cheers,

jeremy.
 



_________________________________________________________________
Amazing prizes every hour with Live Search Big Snap
http://www.bigsnapsearch.com

I am please to announce that v1.06 of Win32::GUI is available for
download from SourceForge and CPAN.

Win32::GUI is a Perl extension allowing creation of native Win32 GUI
applications.

MORE INFORMATION

  Project Homepage:
  http://perl-win32-gui.sourceforge.net/

  Project summary:
  http://sourceforge.net/projects/perl-win32-gui/

  Downloads:
  - Source and ActiveState Perl PPM distributions:
    http://sourceforge.net/project/showfiles.php?group_id=16572
  - Source only:
    http://search.cpan.org/~robertmay/Win32-GUI-1.06/

  Release notes:
  https://sourceforge.net/project/shownotes.php?release_id=576262

RELEASE NOTES AND CHANGES

NAME
    Win32::GUI::ReleaseNotes::RN_1_06 - release notes for v1.06 of
    Win32::GUI

Release Date
    13th February, 2008.

Summary of Changes
    This is a summary of changes between V1.05 and V1.06 See the CHANGELOG
    file in the distribution for the full detail.

    This release is a primarily a bug-fix release addressing the issues
    below. There are a small number of new features. Minor changes have been
    made to enable correct building with the new perl 5.10.0

    Please note that this is intended to be the last release that is
    compatible with the Perl 5.6 series.

  New Features
    More constants
        Win32::GUI::Constants now has over 2000 constants. Please keep
        reporting any that you need that are missing.

    Win32::GUI::DC::DrawFrameControl can DFCS_ADJUSTRECT
        DrawFrameControl() will now adjust it's input parameters if the
        DFCS_ADJUSTRECT flag is set and the input parameters are not
        readonly.

    Dialog navigation for MDI Windows
        Salvadore Oritz submitted a patch to allow 'TAB' navigation between
        controls of an MDI child window. The '-dialogui' is now functional
        on MDI child windows.

    Confusion with GetParent() and TreeView windows
        The GetParent() method usually gets the parent window, but for the
        Treeview class it gets the parent node:

          $parent_node = $tree_view->GetParent($node);

        This made it difficult to get the parent window. Now, if the
        GetParent() method is called without an arguements it returns the
        parent window:

          $parent_window = $tree_view->GetParent();

  Bug Fixes
    Balloon Tooltip info and warning icons swapped
        The info and warning icon for balloon tooltips were the wrong way
        round. This is fixed. (Tracker: 1709017)

    Textfield::GetLine() truncates lines
        Textfield::GetLine() has been re-written to stop it truncating
        lines, and to correctly return empty lines.

    '-prompt' option to Textfield
        The '-prompt' option to the Textfield constructor would generate
        warnings and place the label incorrectly when used to supply a label
        only:

          -prompt => 'Some Label:',

        This should now be fixed. Note that if you were relying on the
        previouly incorrect placement (always against the left border), then
        you may need to re-position your labels. (Tracker: 1734697)

    Warning During Global Destruction
        It was common to get a warning like:

          (in cleanup) Can't call method "DELETE" on an undefined
value at C:/Perl
          /site/lib/Win32/GUI.pm line 3451 during global destruction.

        This is now, hopefully, fixed. Please report if you still see errors
        like this.

    Memory leaks
        Various memory leaks have been resolved.

    Win32::GUI::Scintilla buffer sizes
        There were several places in the Scintilla wrapper code where
        buffers of the wrong sizes were allocated. This has been fixed.

    Win32::GUI::Scroll return value
        On failure Win32::GUI::Scroll was supposed to return undef, but was
        returning random garbage. This has been fixed, along with updates to
        the documentation.

    Win32::GUI::RichEdit::LimitText()
        There was a typo in the function name (LimiteText()). Both spellings
        are now supported.

    ListView BeginLabelEdit event
        The second argument was being set incorrectly when the callback was
        made. This is now fixed. (Tracker: 1706901)

    ListView SelectedItems may now return an empty list
        Win32::GUI::ListView::SelectedItems used to return "undef" when
        there were no items selected. It now returns an empty list, to
        better support things like:

          for my $item ($listview->SelectedItems()) {
            # do something with the selected $item
          }

    TreeView::GetItem may now return an empty list
        Win32::GUI::TreeView::GetItem (and Win32::GUI::TreeView::ItemInfo)
        used to return "undef" when asked about a non-existing item. In now
        returns an empty list to better support things like:

          my %node_info = $treeview->GetItem($item);

    Crash when destroying a window during a callback
        It is possible to write code that destroys a window during one of
        it's one callbacks. This could crash perl. Hopefully this is now
        fixed.

    All Win32::GUI::DC::Poly* drawing functions broken
        Polygon(), PolyBezier(), PolyBezierTo(), PolyLine() and PolyLineTo()
        now all work.

    Comboxbox constructor - using multiple types
        The Win32::GUI::Combobox constructor has been updated to handle the
        '-simple', '-dropdown' and '-dropdownlist' options more intuatively.
        The docuentation has been updated to explain that only one of these
        options should be used, and what happens if more than one is used.

    Win32::GUI::Imagelist::AddMasked broken
        This call should now work. (Tracker: 1734697)

    Win32::GUI::Acceptfiles() generates warning
        The AcceptFiles() method was using some undefined constants. They
        have been removed.

    Build process fixes
        A number of warnings when building under cygwin were fixed - patches
        submitted by Reini Urban.

    Win32::GUI::BitmapInline wasn't using a suitable temporary directory
        Win32::GUI::BitmapInline was always using the current directory to
        create temporary files. This failed when the current directory
        wasn't writable (likely under limited user accounts in Win2K and
        above). It now uses "File::Spec->tmpdir()" to get a writable
        temporary directory. (Tracker: 1586643)

        Further, the filename chosen wasn't thread-safe. This (unlikely)
        race-condition should also be fixed.

    Toolbar Addbitmap can't be called multiple times
        The Win32::GUI::Toolbar::AddBitmap() method failed when called a
        second time, reporting that you couldn't use it when an imagelist
        was already assigned (regardless of whether you had assigned an
        imagelist or not). It may now be called more than once to add
        individual bitmaps to the toolbar.

Deprecated feature status
    This section documents features that have been deprecated in this
    release, or in recent releases, and feature that will be deprecated in
    up-coming releases.

  Win32::GUI::Constants
    The introduction of Win32::GUI::Constants in v1.04 means that we now
    have access to a very large number of constants, so the v1.03 behaviour
    of Win32::GUI to export all constants to the calling namespace by
    default is no longer appropriate. So, a bare

      use Win32::GUI;

    now generates a warning that the old default behaviour will be
    deprecated - although the export behaviour of Win32::GUI v1.03 is
    maintained except for this warning.

    To eliminate this warning and correct your script, do one of the
    following:

    If you don't need any constants, use the empty list:
          use Win32::GUI();

    If you need some constants, name them explicitly:
          use Win32::GUI qw(ES_WANTRETURN CW_USEDEFAULT); # Two
constants exported
          use Win32::GUI qw(/^MB_/);   # Export all constants starting with MB_

    See the Win32::GUI::Constants documentation for the full allowable
    syntax.

    You are advised to fix your scripts now, as a future version will stop
    exporting any constants by default.

    Although not advised, you can suppress the warnings by turning
    deprecated warnings off:

      no warnings 'deprecated';

    Additionally accessing constants from within the Win32::GUI namespace is
    deprecated. I.e.

       -addstyle => Win32::GUI::WS_BORDER,

    will generate a warning with this release, and will stop working with a
    future release. Use one of the following methods instead:

    use the Win32::GUI::Constants namespace instead
          -addstyle => Win32::GUI::Constants::WS_BORDER,

    use any other namespace you fancy
          use Win32::GUI qw(-exportpkg => A::B -autoload);
          ...
          -addstyle => A::B::WS_BORDER,

    maintain compatibility of existing scripts
          use Win32::GUI::Constants qw(-exportpkg => Win32::GUI
:compatibility_win32_gui);
          ...
          -addstyle => Win32::GUI::WS_BORDER,

  Win32::GUI::NotifyIcon
    It is no longer necessary to use the '-id' option to any of the
    Win32::GUI::NotifyIcon methods. The ID is now entirely handled
    internally. You will receive deprecated warnings if you use it.

    In particular, removing Icons from the system tray should be done using

      $NI->Remove();

    and not by the (now deprecated)

      $NI->Delete(-id => 1);

    Use of the "-id" option will generate a warning.

Contributors to this release
    Robert May
    Brian Millham
    Glenn Munroe
    George
    Glenn Linderman
    Salvadore Oritz
    Joseph Cordero
    Reini Urban
    Jeremy White

On 09/02/2008, Jeremy White <jez...@ho...> wrote:
> Done some more testing - had an issue where I was manually removing the linefeed
> characters after this method was called - this resulted in the text being truncated
> by two chars. Changed my code and it seems fine.


So it seems like all the reported issues have been dealt with.  I'll
give it a couple more days and then make a 1.06 release.

Rob.

>> Yep, should be. Once the fix is in I will have to do more testing
>> as the scintillia issue was blocking some functionality - but it looks good:)
>
> Fix is in CVS now. Thanks for your time.
> Rob.

Done some more testing - had an issue where I was manually removing the linefeed characters after this method was called - this resulted in the text being truncated by two chars. Changed my code and it seems fine.

Thanks for committing the fix.

Cheers,

jez.

_________________________________________________________________
Share what Santa brought you
https://www.mycooluncool.com

On 09/02/2008, Jeremy White <jez...@ho...> wrote:
>> Typo. That should be SendMessage rather SendMessageNN.
>>
>>  Is that all your issues dealt with?
>
> Yep, should be. Once the fix is in I will have to do more testing
> as the scintillia issue was blocking some functionality - but it looks good:)

Fix is in CVS now.  Thanks for your time.
Rob.

> Typo. That should be SendMessage rather SendMessageNN.
>
> Fix on its way to CVS now.
>
> Many thanks for taking the time to investigate this. Is that all your
> issues dealt with?

Yep, should be. Once the fix is in I will have to do more testing as the scintillia issue was blocking some functionality - but it looks good:)

Cheers,

jez.


_________________________________________________________________
Telly addicts unite!
http://www.searchgamesbox.com/tvtown.shtml

On 09/02/2008, Jeremy White <jez...@ho...> wrote:
> Ok - found the problem:
>
> Scintilla.PL line 466:
>
> print '  my $lenght = $self->GetTextLength() + 1;', "\n";
>
> changes to:
>
> print '  my $length = $self->SendMessageNN(2182, 0, 0); # includes trailing NUL', "\n";
>
> But the method SendMessageNN doesn't exists in the XS module.
>
> Not to sure what the fix should be - a typo in the send message method name
> or should there be a method called SendMessageNN in the XS module?

Typo. That should be SendMessage rather SendMessageNN.

Fix on its way to CVS now.

Many thanks for taking the time to investigate this.  Is that all your
issues dealt with?

Regards,
Rob.

> Still have the scintilla issues - I should be able to do some debugging this weekend.
>

Ok - found the problem:


http://perl-win32-gui.cvs.sourceforge.net/perl-win32-gui/Win32-GUI/Win32-GUI-Scintilla/Scintilla.PL?r1=1.5&r2=1.6

Scroll down to line 466:

print '  my $lenght = $self->GetTextLength() + 1;', "\n";

changes to:

print '  my $length = $self->SendMessageNN(2182, 0, 0); # includes trailing NUL', "\n";

But the method SendMessageNN doesn't exists in the XS module.

Not to sure what the fix should be - a typo in the send message method name or should there be a method called SendMessageNN in the XS module?

Cheers,

jez. 
_________________________________________________________________
Free games, great prizes - get gaming at Gamesbox. 
http://www.searchgamesbox.com

> Please could those who reported problems with RC1 download again, and
> confirm that their problems are resolved.
>
> New reports, successful or otherwise, are welcome from anyone - please
> reply to this thread.

The following errors have been fixed:

Can't call method "FETCH" on an undefined value during global destruction.
Callback called exit during global destruction.

Still have the scintilla issues - I should be able to do some debugging this weekend.

Cheers,

Jeremy.

_________________________________________________________________
Telly addicts unite!
http://www.searchgamesbox.com/tvtown.shtml

I have made a second release candidate for V1.06 (1.06 RC2 aka
v1.05_91) available for download from:

http://www.robmay.me.uk/win32-gui-1-06-rc2

This release candidate (hopefully) fixes a number of issues that were
reported with RC1:

- Perl 5.8 PPM failed with perls earlier than 5.8.8
- 'Use of undefined value in subroutine entry' when using the WS_POPUP style
- 'Use of undefined value in subroutine entry (during global
destruction)' with Scintialla.

Please could those who reported problems with RC1 download again, and
confirm that their problems are resolved.

New reports, successful or otherwise, are welcome from anyone - please
reply to this thread.

We're getting very close to a V1.06 final release.

Regards,
Rob.

On 31/01/2008, Jeremy White <jez...@ho...> wrote:
>
> Just testing the latest Win32-GUI code line and I'm getting a few odd errors
> (mingw & 5.8.8) It could be my environment and I haven't got time to dig deeper,
> but thought I'd report them just in case it's something simple.
>
> When I exit my app I get lots error messages:

> Can't call method "FETCH" on an undefined value during global destruction.
> ...
> Can't call method "FETCH" on an undefined value during global destruction.

> With the final line being:
> Callback called exit during global destruction.

I can't reproduce this one yet.

> Also I'm using Scintilla ... I get the following error:
>
> Can't locate auto/Win32/GUI/Scintilla/SendMessage.al in
> @INC (@INC contains C:/Perl/site/lib C:/Perl/lib .) at
> C:/Perl/site/lib/Win32/GUI/Scintilla.pm at line 1430

Nor this one.

I have just found a problem that causes:

Use of uninitialized value in subroutine entry at C:\...\blib\lib/Win32/GUI.pm
line 3480 during global destruction.

Fix on it's way into CVS now - I wonder if that is the underlying
cause??  This error was being raised by one of the Scintilla demos and
one of the DIBitmap demos.

Sadly I can't find a repeatable, short example to turn into a test case.

Regards,
Rob.

>> In my main development environment (Vista, 5.8.8 (822)) I did the same (uninstalled and
>> used the binary) and am getting the same issues (undefined value during destruction and
>> Scintillia) I mentioned a few days ago.
>
> Hmmm. I wonder what's causing that. I had really hoped that I had
> resolved those destruction warnings. Are these warnings that you don't
> get with Win32::GUI v1.05?

Yes, I dont get them with v1.05. The only thing that I can think of is that I use a lot of child windows and for some reason they are getting destoryed twice somehow?

Any thoughts on the scintilla issue?

I'll try and find time to debug this week - although I need to test the latest version of PerlApp against Vista!

Cheers,

jez. 
_________________________________________________________________
Free games, great prizes - get gaming at Gamesbox. 
http://www.searchgamesbox.com

On 05/02/2008, Jeremy White <jez...@ho...> wrote:
>
> Was unable to do any testing over the weekend. Sorry.

Don't worry.

> I uninstalled the previous
> version (via ppm) and installed the binary for 5.8 on my "clean" build machine:
>
> XP, 5.8.7 (813):
>
> Running any win32::GUI script I get:
>
> A windows error message box: "The procedure entry point Perl_sv_2uv_flags could not be located in the dynamic link library perl58.dll"
>
> Once you click on, you get the following on the command line.
>
> C:\DevelopmentVPN\Win32-GUI-1.05_90-PPM-5.8>win32-gui-demos
> Can't load 'C:/Perl/site/lib/auto/Win32/GUI/GUI.dll' for module Win32::GUI: load
> _file:The specified procedure could not be found at C:/Perl/lib/DynaLoader.pm li
> ne 230.
>  at C:\Perl\bin\win32-gui-demos.bat line 42
> Compilation failed in require at C:\Perl\bin\win32-gui-demos.bat line 42.
> BEGIN failed--compilation aborted at C:\Perl\bin\win32-gui-demos.bat line 42.

Oops.  That's my fault.  RC1 PPM for Perl 5.8 was built against perl
5.8.8 not perl 5.8.0, so will have backwards compatibility issues.  If
I get cause to generate a new RC I'll fix it then, otherwise I'll fix
it in the actual release.

> In my main development environment (Vista, 5.8.8 (822)) I did the same (uninstalled and
> used the binary) and am getting the same issues (undefined value during destruction and
> Scintillia) I mentioned a few days ago.

Hmmm.  I wonder what's causing that.  I had really hoped that I had
resolved those destruction warnings. Are these warnings that you don't
get with Win32::GUI v1.05?

Thank you for taking the time to do this much.

Regards,
Rob.

> If you download it please feedback your experience (even if it is only
> 'OK so far') by replying to this thread (or directly to me) so that I
> can get a feel for whether there are any significant issues with the
> release.

Was unable to do any testing over the weekend. Sorry. I uninstalled the previous version (via ppm) and installed the binary for 5.8 on my "clean" build machine:

XP, 5.8.7 (813):

Running any win32::GUI script I get:

A windows error message box: "The procedure entry point Perl_sv_2uv_flags could not be located in the dynamic link library perl58.dll"

Once you click on, you get the following on the command line.

C:\DevelopmentVPN\Win32-GUI-1.05_90-PPM-5.8>win32-gui-demos
Can't load 'C:/Perl/site/lib/auto/Win32/GUI/GUI.dll' for module Win32::GUI: load
_file:The specified procedure could not be found at C:/Perl/lib/DynaLoader.pm li
ne 230.
 at C:\Perl\bin\win32-gui-demos.bat line 42
Compilation failed in require at C:\Perl\bin\win32-gui-demos.bat line 42.
BEGIN failed--compilation aborted at C:\Perl\bin\win32-gui-demos.bat line 42.

C:\DevelopmentVPN\Win32-GUI-1.05_90-PPM-5.8>

In my main development environment (Vista, 5.8.8 (822)) I did the same (uninstalled and used the binary) and am getting the same issues (undefined value during destruction and Scintillia) I mentioned a few days ago.

I am snowed under at the moment so I wont be able to create example showing the errors or to provide a fix. If you don't get anyone else reporting these issues - just release the version anyway - I am doing heavy things with Win32-GUI so it could even be my code.

Sorry for the useless bug reports!

Cheers,

jez.


_________________________________________________________________
Get Hotmail on your mobile, text MSN to 63463!
http://mobile.uk.msn.com/pc/mail.aspx

All,

I have made source and binary[1] distributions of the first release
candidate available for download.

Win32::GUI v1.05_90 (aka v1.06 RC1) can be found here:
http://www.robmay.me.uk-a.googlepages.com/win32-gui-1-06-rc1

This download page cannot cope with high volume downloads, is
temporary and may disappear at any time.

If you download it please feedback your experience (even if it is only
'OK so far') by replying to this thread (or directly to me) so that I
can get a feel for whether there are any significant issues with the
release.

In the absence of any major issues I hope to turn this into a final
1.06 release in about a week's time.

Regards,
Rob.

[1] PPM binary distributions for ActiveState Perl versions 5.6.x,
5.8.x and 5.10.x

On 01/02/2008, Robert May <ro...@th...> wrote:
> On 01/02/2008, Robert May <ro...@th...> wrote:
> > > > - Find out why my build on Cygwin results in an unloadable
> > > > SciLexer.dll (permissions I think)
> > >
> > > Yes. On cygwin it is missing the x "permission", which is "Read &
> > > Execute" on Windows ACLs.
> >
> > I'll investigate further.
>
> I just raised a support request to add the execute bit to SciLexer.dll
> in the repository.

It's been done - although it may take "some time" to propagate.

Rob.

On 01/02/2008, Reini Urban <ru...@x-...> wrote:
> 2008/2/1, Robert May <rob...@us...>:
> > I can now get clean build/test in the following environment:
> >
> > OS  :   MSWin32      MSWin32      MSWin32           Cygwin
> > Perl:   AS5.6.1(638) AS5.8.8(822) AS5.10.0(1002)    5.8.8
> >
> > VC6     ok           ok           ok                -
> > gcc      ok           ok           ok                ok[1]
> >
> > [1] After I correct the SciLexer.dll permissions
> >
> > I'd really like the following (Reini?):
> > Confirmation of build/test on Cygwin, perl 5.8.8
> > Test/build on Cygwin, perl 5.10.0
>
> Confirmation done.
> Both perl's build and test ok.

Many thanks.  I'll get a release candidate up somewhere this weekend.

Rob.

2008/2/1, Robert May <rob...@us...>:
> I can now get clean build/test in the following environment:
>
> OS  :   MSWin32      MSWin32      MSWin32           Cygwin
> Perl:   AS5.6.1(638) AS5.8.8(822) AS5.10.0(1002)    5.8.8
>
> VC6     ok           ok           ok                -
> gcc      ok           ok           ok                ok[1]
>
> [1] After I correct the SciLexer.dll permissions
>
> I'd really like the following (Reini?):
> Confirmation of build/test on Cygwin, perl 5.8.8
> Test/build on Cygwin, perl 5.10.0

Confirmation done.
Both perl's build and test ok.

> I need to work with Jeremy to find out where his problems are coming
> from (Vista?)

Don't let this delay releasing the build to the users list - it could just =
be me:)

Cheers,

jez.

_________________________________________________________________
Free games, great prizes - get gaming at Gamesbox.=20
http://www.searchgamesbox.com=

On 31/01/2008, Robert May <ro...@th...> wrote:
> Here's what I have left to do before a 1.06 release:
>
> - Check that some 'redefined definitions' warnings have been removed
> by the updates to ppport.h

Appears good.

> - Find out why my build on Cygwin results in an unloadable
> SciLexer.dll (permissions I think)

Awaiting SourceForge Admins to add permissions.  Release could happen
without this - I just need to remember to make the source tarball have
the execute bit set on SciLexer.dll

> - Add to all Makefiles a suitable addition of the -mms-bitfields flag
> for gcc, to allow building under 5.10 with GCC against a VC++ compiled
> Perl to work.

Done

> - check that the new tests for Scintilla work under cygwin

Done

> - update $VERSION of modules that have changes

Done

> - update Win23::GUI $VERSION to 1.05_99 and make a beta distribution available.

Done


I can now get clean build/test in the following environment:

OS  :   MSWin32      MSWin32      MSWin32           Cygwin
Perl:   AS5.6.1(638) AS5.8.8(822) AS5.10.0(1002)    5.8.8

VC6     ok           ok           ok                -
gcc     ok           ok           ok                ok[1]

[1] After I correct the SciLexer.dll permissions

I'd really like the following (Reini?):
Confirmation of build/test on Cygwin, perl 5.8.8
Test/build on Cygwin, perl 5.10.0

I need to work with Jeremy to find out where his problems are coming
from (Vista?)

Nearly there,
Rob.

> Note the other cygwin permission problems:
> * GUI.rc is not writable, so the default make will fail.

Hmmm.  It's not read-only in the repository or in my build directory.
I did (somehow) get a read-only version in the 1.05 source tarball -
is that what you're talking about?

On the basis on my investigation I don't think this is a real problem.

> * samples/* are not writable, so the blib update for the demos will fail.
> No patch, I fix that in my build script.

On further investigation I don't know what you refering to here
either.  The samples get put into blib just fine.  There's a problem
with the '*' glob picking up the CVS directory, but that's ignored
(although I'll fix it).

Regards,
Rob.