202751 - nsBlender can have issues on 24-bit GTK setups

Status: RESOLVED FIXED

(Core Graveyard :: GFX, defect, P2)

Description

[Boris Zbarsky [:bzbarsky]]

The basic problem is that the nsDrawingSurfaceGTK::Lock sets the drawing surface
width in bytes as:

177   *aWidthBytes = aWidth*(mImage->bpp >> 3);

and mImage->bpp is actually 32, not 24, on 24-bit setups.

The following patch makes the octopus not crash (not suitable for checkin; needs
a zero-check at the very least):

Index: gfx/src/nsBlender.cpp
===================================================================
RCS file: /cvsroot/mozilla/gfx/src/nsBlender.cpp,v
retrieving revision 3.35
diff -p -u -2 -r3.35 nsBlender.cpp
--- gfx/src/nsBlender.cpp       6 Apr 2003 02:49:30 -0000       3.35
+++ gfx/src/nsBlender.cpp       21 Apr 2003 04:46:49 -0000
@@ -383,6 +383,5 @@ NS_IMETHODIMP nsBlender::GetAlphas(const
         *aAlphas = new PRUint8[r.width*r.height];
         if (*aAlphas) {
-          PRUint32 depth;
-          mContext->GetDepth(depth);
+          PRUint32 depth = (blackBytesPerLine/r.width)*8;
           ComputeAlphas(r.height, blackBytesPerLine, depth,
                         blackBytes, whiteBytes, blackSpan,

Comment 1

[Boris Zbarsky [:bzbarsky]]

This is a crash, actually.... ;)

Comment 2

[Thomas Thurman]

The patch seems to stop the crash for me, but makes the window appear kind of
transparent instead. You can see right through to the window underneath where
it's supposed to be white.

Comment 3

[Thomas Thurman]

Sorry, please ignore my last comment-- now that I can see
the window under Phoenix, I find that the CSS was
subtly broken, hence the transparency.

Comment 4

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Attachment: fixes

OK, this fixes the crash ... it's just the bz patch. It also restores
transparent chrome functionality --- the window is set to transparent before it
is shown, but it isn't painted before it is shown, so the alpha mask bitmap is
never setup. We should just test the transparency state directly.

Comment 5

[Christopher Blizzard (:blizzard)]

Comment on attachment 130754 [details] [diff] [review]
fixes

What's with the PAINT_SEPERATELY in this patch?

Comment 6

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Attachment: better patch

that patch was bogus. Check this out.

Comment 7

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Fix checked in.