Closed Bug 581083 Opened 14 years ago Closed 14 years ago

REFTEST TEST-UNEXPECTED-FAIL | file:///c:/talos-slave/mozilla-central-win7-opt-u-reftest-d2d/build/reftest/tests/layout/reftests/svg/dynamic-small-object-scaled-up-02.svg |

Categories

(Core :: Graphics, defect)

x86
Windows Vista
defect
Not set
normal

Tracking

()

RESOLVED FIXED
Tracking Status
blocking2.0 --- betaN+

People

(Reporter: jrmuizel, Assigned: bas.schouten)

References

Details

Attachments

(1 file)

This one has a pretty substantial difference.
Assignee: nobody → bas.schouten
Status: NEW → ASSIGNED
OS: Mac OS X → Windows Vista
The problem here is that we're inverse transforming the stroke in order to set the transform on the rasterizer. This is needed because the miter angles and stroke width and such are untransformed. Because this path was really small when inverse transformed back to its original, the cairo fixed point precision of 8 fractional bits was insufficient to store the stroke properly and rounding errors occurred. I know inverse the transformation in D2D which uses floating point, avoiding the issue. This fixes the reftest.
Attachment #460186 - Flags: review?(jmuizelaar)
Comment on attachment 460186 [details] [diff] [review]
Fix transformed stroke drawing

Why do the box tests go away?
(In reply to comment #2)
> Comment on attachment 460186 [details] [diff] [review]
> Fix transformed stroke drawing
> 
> Why do the box tests go away?

Because I'm drawing a transformed geometry anyway. I could transform a rectangle geometry but I doubt it would perform that much fast, I could bring it back though.
Comment on attachment 460186 [details] [diff] [review]
Fix transformed stroke drawing

Make sure you get rid of the DEBUG_LEVEL changes.
Attachment #460186 - Flags: review?(jmuizelaar) → review+
This needs blocking to land.
blocking2.0: --- → ?
Blocks a blocker, let's go.
blocking2.0: ? → betaN+
Pushed http://hg.mozilla.org/mozilla-central/rev/d29e7fff79b6.
Status: ASSIGNED → RESOLVED
Closed: 14 years ago
Resolution: --- → FIXED
This or the gradient patch seems to have caused a bunch of new reftest failures. Bas, can you try to figure out which and why?
I believe these are intermittent failures sadly, I'm not sure why, when I tested locally I looked very briefly at a few, but they weren't really rendering problems. It was like two textboxes showing completely different text and such things. But I didn't look at what seemed to cause it yet.
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: