Folks,
This is weird one. The Winelib build of AbiWord exposes an off-by-one error. However, the Win32 build of AbiWord works perfectly (without this very problem) under Wine/ReactOS/Windows. AbiWord builds flawlessly under Winelib, not even a warning, so my guess one of our constants is off by one. The bug must appear in their code while building it, since our code to execute the Winelib app and the Win32 app is byte for byte identical.
I've attached a few screenshots to demonstrate what happens when you open a menu/dialog/etc and then you close it. Namely, the invalidation (or rather painting) is off by one at right & bottom of the invalidated rectangle.
On Sun, 18 Apr 2004 12:23:11 -0400, Dimitrie O. Paun wrote:
This is weird one. The Winelib build of AbiWord exposes an off-by-one error. However, the Win32 build of AbiWord works perfectly (without this very problem) under Wine/ReactOS/Windows. AbiWord builds flawlessly under Winelib, not even a warning, so my guess one of our constants is off by one. The bug must appear in their code while building it, since our code to execute the Winelib app and the Win32 app is byte for byte identical.
Can you build on Windows with our headers, and maybe substitute them one at a time until you locate the header of ours which is wrong?
On April 18, 2004 6:12 pm, Mike Hearn wrote:
Can you build on Windows with our headers, and maybe substitute them one at a time until you locate the header of ours which is wrong?
Way too time consuming for me, I'm rather tied up ATM. If anyone wants to look into this problem, I can help get them started...
From: Mike Hearn
On Sun, 18 Apr 2004 12:23:11 -0400, Dimitrie O. Paun wrote:
This is weird one. The Winelib build of AbiWord exposes an off-by-one error. However, the Win32 build of AbiWord works perfectly (without this very problem) under Wine/ReactOS/Windows. AbiWord builds flawlessly under Winelib, not even a warning, so my guess one of our constants is off by one. The bug must appear in their code while building it, since our code to execute the Winelib app and the Win32 app is byte for byte
identical.
Can you build on Windows with our headers, and maybe substitute them one at a time until you locate the header of ours which is wrong?
I tried it the other way around, copied the w32api headers to my /usr/local/include/wine/windows directory and rebuilt the Winelib AbiWord. The problem is still there....
Gé van Geldorp.
On Sunday 18 April 2004 18:23, Dimitrie O. Paun wrote:
Folks,
This is weird one. The Winelib build of AbiWord exposes an off-by-one error. However, the Win32 build of AbiWord works perfectly (without this very problem) under Wine/ReactOS/Windows. AbiWord builds flawlessly under Winelib, not even a warning, so my guess one of our constants is off by one. The bug must appear in their code while building it, since our code to execute the Winelib app and the Win32 app is byte for byte identical.
I've attached a few screenshots to demonstrate what happens when you open a menu/dialog/etc and then you close it. Namely, the invalidation (or rather painting) is off by one at right & bottom of the invalidated rectangle.
from wine/programs/winemine/README "Starting a new game causes the window to drop one pixel"
And it still do on my wine-cvs build but not then I test the windows builds downloaded from SourceForge.net on win98. If this problem is related could it be simpler to take a look on winemine then on AbiWord.
Acke
-
Acke Carlsson -
Dimitrie O. Paun -
Ge van Geldorp -
Mike Hearn