Open Bug 679257 Opened 14 years ago Updated 3 years ago

EXTEND_NONE is used instead of EXTEND_PAD when src sample area is entirely within the surface

Categories

(Core :: Graphics, defect)

Product:
Core
Component:
Graphics
Platform:
x86
Linux
Type:
defect

Tracking

()

People

(Reporter: fry.kun, Unassigned)

References

(
URL
)

Details

Example: http://demos.hacks.mozilla.org/openweb/HWACCEL/ Demo appears correct but runs slowly because it falls back to software rendering. I've verified with debug tracing that REPEAT_NONE is still used for that test case. I thought Bug 468496 would fix it, but apparently it deals with a different test case
Behavior still present in 7.0a2
URL: http://demos.hacks.mozilla.org/openwe...
Depends on: 468496
I would have guessed changes in bug 468496 and bug 620216 were enough to avoid RepeatNone on the HWACCEL test. The speed of the demo on my system with radeon ddx seems to suggest so. Where did you see RepeatNone in use? Are you sure this is not the source image size slowing things down (bug 620065 comment 19)?
Yes, I've altered the same demo to use a smaller image (small enough to work with my hardware), but as long as it's left as a jpeg I keep getting this fallback. If I convert to a .png, it works correctly.
What is the debug tracing you refer to?
No need to answer comment 4. _cairo_pattern_acquire_surface_for_surface is the culprit: http://cgit.freedesktop.org/cairo/tree/src/cairo-pattern.c?id=ae2b7b13cd5fdeaee44496056bb99f497346e262#n2428
Status: UNCONFIRMED → NEW
Ever confirmed: true
Summary: EXTEND_NONE is used instead of EXTEND_PAD in canvas..drawImage → EXTEND_NONE is used instead of EXTEND_PAD when src sample area is entirely within the surface
Does that mean I can try a fixed version some time real soon? :D
Well, it's easy to remove the "trim" code from cairo to try, but it might be debatable whether the right fix is in cairo or the X server or DDX. It would be interesting to know why that code is in cairo, and probably worth a cairo bug report - the code looks unnecessary, and it would be easier to fix cairo than the various drivers. (https://bugs.freedesktop.org/show_bug.cgi?id=35579#c27)
Great, he says it's a driver bug. Could you please communicate directly, it's the same bugzilla, for crying out loud! And remember, "neither party has a bug" is a wrong answer!
http://www.mikematas.com/preventi/ is a "night and day" testcase.
Severity: normal → S3
You need to log in before you can comment on or make changes to this bug.