[Bug 20164] New: Regression : Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

Summary: Regression : Icons incorrectly rendered in RSTAB
           Product: Wine
           Version: 1.1.30
          Platform: PC
        OS/Version: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: winex11.drv
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]
                CC: [e-mail]

Created an attachment (id=23761)
 --> (http://bugs.winehq.org/attachment.cgi?id=23761)
Bad (above) and correct (below) behaviour of icons in RSTAB toolbar

this patch :

http://source.winehq.org/git/wine.git/?a=commit;h=004eabfbd398192c0de92bafe1...

caused a regression in RSTAB application, as shown by attached image.

Commenting out XRender code here (in winex11.drv/bitblt.c around line 1440

[code]
    if (useSrc)
    {
        wine_tsx11_lock();
        pixmaps[SRC] = XCreatePixmap( gdi_display, root_window, width, height,
                                      physDevDst->depth );
        wine_tsx11_unlock();

//        if(!X11DRV_XRender_GetSrcAreaStretch( physDevSrc, physDevDst,
pixmaps[SRC], tmpGC,
//                                              widthSrc, heightSrc, widthDst,
heightDst,
//                                              &visRectSrc, &visRectDst))
//        {
            if (fStretch)
                BITBLT_GetSrcAreaStretch( physDevSrc, physDevDst, pixmaps[SRC],
tmpGC,
                                          xSrc, ySrc, widthSrc, heightSrc,
                                          xDst, yDst, widthDst, heightDst,
                                          &visRectSrc, &visRectDst );
            else
                BITBLT_GetSrcArea( physDevSrc, physDevDst, pixmaps[SRC], tmpGC,
                                  &visRectSrc );
//        }
    }

[/code]

makes things working again.

Ciao

Max

[Bug 20164] Regression : Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

--- Comment #1 from Roderick Colenbrander   2009-09-26 07:07:26 ---
Could you try if this patch which I need to resubmit fixes the issue:
http://www.winehq.org/pipermail/wine-patches/2009-September/078446.html ?

[Bug 20164] Regression : Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

--- Comment #2 from [e-mail]  2009-09-26 07:18:31 ---
(In reply to comment #1)
> Could you try if this patch which I need to resubmit fixes the issue:
> http://www.winehq.org/pipermail/wine-patches/2009-September/078446.html ?

Sorry, your patch don't apply cleanly to current git, and I've got no time by
now to fix it :-(
Let me know if you rebase it to 1.1.30.

Ciao

Max

[Bug 20164] Regression : Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

--- Comment #3 from Roderick Colenbrander   2009-09-26 07:54:57 ---
Created an attachment (id=23769)
 --> (http://bugs.winehq.org/attachment.cgi?id=23769)
Possible fix

Try this patch.

Roderick

[Bug 20164] Regression : Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

Vitaliy Margolen  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |regression
           Severity|normal                      |trivial

[Bug 20164] Regression : Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |patch

[Bug 20164] Regression : Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

[e-mail] changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|trivial                     |normal

--- Comment #4 from [e-mail]  2009-09-26 16:17:21 ---
Nope, the patch doesn't fix the issue.
The icons remembers me something when I was developing my dib engine, and it
was related to mixed blitting between dib engine managed bitmaps and x11drv
managed ones.....

Changing importance again to normal, I don't see anything trivial on that one.

Max

[Bug 20164] Regression : Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

--- Comment #5 from Roderick Colenbrander   2009-09-26 17:09:02 ---
What display drivers are you using? Basically I'm using XRender in this case
for copying the data instead of plain xlib calls. It can also perform
stretching and other operations. Perhaps it is something related to the 1-bit
masking code.

You could try to look at the xrender function. You'll see two codepaths:
    /* mono -> color */
    if(physDevSrc->depth == 1)
    ...
    else /* color -> color but with different depths */
    ...

First figure out in which codepath it goes wrong e.g. by just returning false
in the codepath, since then we will fall back to the old non-xrender code. If
it is in the 1-bit code perhaps the wrong background color is used or so. Did
this regression occur in 1.1.30 or did it occur earlier as well? Since I added
X11DRV_XRender_GetSrcAreaStretch a couple of release ago.

Thanks,
Roderick

[Bug 20164] Regression : Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

--- Comment #6 from [e-mail]  2009-09-26 17:28:57 ---
(In reply to comment #5)
> ...... Did
> this regression occur in 1.1.30 or did it occur earlier as well? Since I added
> X11DRV_XRender_GetSrcAreaStretch a couple of release ago.
>
> Thanks,
> Roderick

Hi Roderick,

the regression occurred exactly on the patch pointed on the bug report. I
didn't notice before because I was using an old build on these days... no time
to test new releases because of my job :-(
Anyways, I just updated, noticed the regression and made a regression testing
from 1.1.19 to 1.1.30, which pointed me to your patch.
I can't test more in depth right now, I hope I'll have more time on next
days...

I stumbled over a similar problem when I was developing the DIB engine using
the old approach of the mixed pysdev pointers some months ago, and (IIRC) it
was due to mixed blt's between dib engine managed stuff and x11drv ones.
If you look at icons, you can see that some of them are just slightly shifted,
others are completely wrong, so I guess your source bitmap used in blitting is
wrong somehow.
My (old) similar problem was because (IIRC) x11drv has an internal PIXMAP copy
of DIB, (in some kind of linked list ? don't remember...) which was not there
on DIBs managed by my engine. On mixed blitting winex11.drv was reading the
(non-existent) pixmap internal copy instead of the true dib content, which was
of course not the desired effect..... BTW, that's the main reason because my
engine is slower than original on mixed blits; it must convert dib to ddb and
pass it to winex11, as using DIB directly doesn't work because of the missing
internal PIXMAP buffer.

Ah, obviously I tested it on vanilla wine, without my engine, to be sure it was
not due to it :-)
About RSTAB, there is a demo on www.dlubal.com, but now it's on 7.xx version
and I'm using the 5.xx one.... don't know if bug is present on new one.

Ciao

Max

[Bug 20164] Regression : Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

Vitaliy Margolen  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |trivial

--- Comment #7 from Vitaliy Margolen   2009-09-26 22:05:51 ---
Trivial - For a UI glitch that doesn't affect running of a program
http://bugs.winehq.org/page.cgi?id=fields.html#bug_severity

[Bug 20164] Regression : Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

Vitaliy Margolen  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|patch                       |

--- Comment #8 from Vitaliy Margolen   2009-09-26 22:06:07 ---
There is no working patch

[Bug 20164] Regression : Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

--- Comment #9 from [e-mail]  2009-09-27 03:29:52 ---
(In reply to comment #7)
> Trivial - For a UI glitch that doesn't affect running of a program
> http://bugs.winehq.org/page.cgi?id=fields.html#bug_severity

Normal      For an application crash or loss of functionality

If you call "ui glitch only" completely scrambled icons that makes the app
unusable (many commands are available only with toolbar) then, ok, is trivial.
Then I'd call 'trivial' also something that makes screen appear completely
black,
program still runs.....

Max

[Bug 20164] Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

Dmitry Timoshkov  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|Regression : Icons          |Icons incorrectly rendered
                   |incorrectly rendered in     |in RSTAB
                   |RSTAB                       |

[Bug 20164] Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

--- Comment #10 from Roderick Colenbrander   2009-09-27 05:48:12 ---
Version 7 doesn't have the issue. If you can please analyze the code a bit and
figure out where it goes wrong. In short all the function does is copying data
from a source pixmap (which has an associated xrender picture) to a
destination. Either something is wrong with masking or perhaps as you say with
offsets.

Roderick

[Bug 20164] Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

--- Comment #11 from [e-mail]  2009-10-13 18:03:05 ---
(In reply to comment #10)
> Version 7 doesn't have the issue. If you can please analyze the code a bit and
> figure out where it goes wrong. In short all the function does is copying data
> from a source pixmap (which has an associated xrender picture) to a
> destination. Either something is wrong with masking or perhaps as you say with
> offsets.
>
> Roderick

Just adding some fixme printouts, it repeats this pattern :

fixme:xrender:X11DRV_XRender_GetSrcAreaStretch ----------------- START
----------------------
trace:xrender:X11DRV_XRender_GetSrcAreaStretch src depth=24 widthSrc=176
heightSrc=15 xSrc=0 ySrc=0
trace:xrender:X11DRV_XRender_GetSrcAreaStretch dst depth=1 widthDst=176
heightDst=15 xDst=0 yDst=0
fixme:xrender:X11DRV_XRender_GetSrcAreaStretch ----------------Depth of no use

-------- SOME USELESS STUFFS -------

fixme:xrender:X11DRV_XRender_GetSrcAreaStretch ----------------- START
----------------------
trace:xrender:X11DRV_XRender_GetSrcAreaStretch src depth=1 widthSrc=16
heightSrc=15 xSrc=0 ySrc=0
trace:xrender:X11DRV_XRender_GetSrcAreaStretch dst depth=24 widthDst=16
heightDst=15 xDst=0 yDst=0
fixme:xrender:X11DRV_XRender_GetSrcAreaStretch ----------------Converting mono
to color
fixme:xrender:X11DRV_XRender_GetSrcAreaStretch ----------------END
trace:xrender:get_xrender_format Returning wxr_format=0
fixme:xrender:X11DRV_XRender_GetSrcAreaStretch ----------------- START
----------------------
trace:xrender:X11DRV_XRender_GetSrcAreaStretch src depth=1 widthSrc=16
heightSrc=15 xSrc=16 ySrc=0
trace:xrender:X11DRV_XRender_GetSrcAreaStretch dst depth=24 widthDst=16
heightDst=15 xDst=16 yDst=0
fixme:xrender:X11DRV_XRender_GetSrcAreaStretch ----------------Converting mono
to color
fixme:xrender:X11DRV_XRender_GetSrcAreaStretch ----------------END
trace:xrender:get_xrender_format Returning wxr_format=0
fixme:xrender:X11DRV_XRender_GetSrcAreaStretch ----------------- START
----------------------
trace:xrender:X11DRV_XRender_GetSrcAreaStretch src depth=1 widthSrc=16
heightSrc=15 xSrc=32 ySrc=0
trace:xrender:X11DRV_XRender_GetSrcAreaStretch dst depth=24 widthDst=16
heightDst=15 xDst=32 yDst=0
fixme:xrender:X11DRV_XRender_GetSrcAreaStretch ----------------Converting mono
to color
fixme:xrender:X11DRV_XRender_GetSrcAreaStretch ----------------END

and so on.... 
So, it seems it takes a very wide bitmap (presumably containing many icons in
an horizontal strip, in 24 bit/color, convert to a 1bpp, and then uses the 1bpp
to output the single icons.
The 24 to 1 conversion is not handled by your code, the 1 to 24 are... maybe
the problem lies there.

Ciao

Max

[Bug 20164] Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

--- Comment #12 from Roderick Colenbrander   2009-10-13 18:28:36 ---
Hi Max,

If you suspect that this 24->1 is the issue, you could make that case fail and
just return FALSE. When FALSE is returned the code reverts to the old
non-xrender code path. If it works then, then you know for certain that it is
related to that specific code path.

Such a test would be useful.

Roderick

[Bug 20164] Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

--- Comment #13 from [e-mail]  2009-10-14 09:27:29 ---
(In reply to comment #12)
> Hi Max,
>
> If you suspect that this 24->1 is the issue, you could make that case fail and
> just return FALSE. When FALSE is returned the code reverts to the old
> non-xrender code path. If it works then, then you know for certain that it is
> related to that specific code path.
>
> Such a test would be useful.
>
> Roderick

Yep, I didn't think about it.... I made the test and it's indeed the problem.
Returning false on the 24->1 conversion renders icons correctly.

Ciao

Max

[Bug 20164] Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

--- Comment #14 from Wylda   2009-11-16 18:30:07 ---
Created an attachment (id=24796)
 --> (http://bugs.winehq.org/attachment.cgi?id=24796)
Incorectly rendered icons

Hi, nearly two months past, nothing changed = i.e. still not fixed in current
git (wine-1.1.33-88-gd51b4e1) and there is another bug 20732, which in comment
#0 contains tiny example application written as a test even with source code.

bisecting showed me commit:

004eabfbd398192c0de92bafe1b31381a07d7a9c is first bad commit
commit 004eabfbd398192c0de92bafe1b31381a07d7a9c
Author: Roderick Colenbrander 
Date:   Thu Jul 9 21:20:12 2009 +0200

winex11: Add an XRender based GetSrcAreaStretch.

:040000 040000 8eb47f001f4635ac2ebb31c451b0bd5b46b11c78
ee9832dcd1175a51ccaca17f92674b3507a1be1d M      dlls

[Bug 20164] Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

Gabriele Moabiti  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

[Bug 20164] Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

Wylda  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #15 from Wylda   2009-11-16 18:52:08 ---

Because the bug is between 1.1.25 and 1.1.26, version should be changed to
1.1.26, shouldn't be?

--private keyword: bisected

[Bug 20164] Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
            Version|1.1.30                      |1.1.26

[Bug 20164] Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

--- Comment #16 from Roderick Colenbrander   2010-02-17 10:33:51 ---
The RSTAB bug might be fixed in GIT now.

[Bug 20164] Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

--- Comment #17 from Nikolay Sivov   2010-05-24 06:17:47 ---
Please test with 1.2-rc1.

[Bug 20164] Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

André H.  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]
    Regression SHA1|                            |004eabfbd398192c0de92bafe1b
                   |                            |31381a07d7a9c

[Bug 20164] Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

Bruno Jesus  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
                 CC|                            |[e-mail]
         Resolution|                            |FIXED

--- Comment #18 from Bruno Jesus  2012-04-07 11:38:32 CDT ---
This bug has been fixed some months ago with several other toolbar/icons
background problem. Retested in wine 1.5.

Needs winetricks mfc40 vb6run mfc42 prior to the program installation.

If anyone can retest and still can reproduce the issue feel free to reopen.

[Bug 20164] Icons incorrectly rendered in RSTAB

http://bugs.winehq.org/show_bug.cgi?id=20164

Alexandre Julliard  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED

--- Comment #19 from Alexandre Julliard  2012-04-13 13:27:46 CDT ---
Closing bugs fixed in 1.5.2.