Bug 128283 – REGRESSION (r163262?): transitions/shorthand-border-transitions.html flakily asserts

• Status: ASSIGNED
• Component: Layout and Rendering

[Alexey Proskuryakov]

transitions/shorthand-border-transitions.html  has started to flakily assert on WebKit2 debug testers:

ASSERTION FAILED: !rect.isEmpty()

Thread 0 Crashed:: Dispatch queue: com.apple.main-thread
0   com.apple.JavaScriptCore      	0x00000001106fd90a WTFCrash + 42 (Assertions.cpp:333)
1   com.apple.WebCore             	0x000000011209d410 WebCore::GraphicsContext::drawRect(WebCore::FloatRect const&) + 128 (GraphicsContextCG.cpp:293)
2   com.apple.WebCore             	0x0000000112e8b0a1 WebCore::RenderObject::drawLineForBoxSide(WebCore::GraphicsContext*, WebCore::LayoutUnit, WebCore::LayoutUnit, WebCore::LayoutUnit, WebCore::LayoutUnit, WebCore::BoxSide, WebCore::Color, WebCore::EBorderStyle, int, int, bool) + 16193 (RenderObject.cpp:937)
3   com.apple.WebCore             	0x0000000112d58723 WebCore::RenderBoxModelObject::paintOneBorderSide(WebCore::GraphicsContext*, WebCore::RenderStyle const*, WebCore::RoundedRect const&, WebCore::RoundedRect const&, WebCore::LayoutRect const&, WebCore::BoxSide, WebCore::BoxSide, WebCore::BoxSide, WebCore::BorderEdge const*, WebCore::Path const*, WebCore::BackgroundBleedAvoidance, bool, bool, bool, WebCore::Color const*) + 2419 (RenderBoxModelObject.cpp:1695)

A lot of similar assertions were introduced in r163262 and fixed in r163264, but this one is still happening. The first occurrence was on r163263, and it happens about 1 time out of 10.

[Alexey Proskuryakov]

This is readily reproducible locally.

[zalan]

I tried to repro it a few times with no success...I'll try again.

[Alexey Proskuryakov]

I was simply running this command, and could reproduce two times out of three:

run-webkit-tests -2 transitions/shorthand-border-transitions.html

Should probably be easier to reproduce with "--iterations 100".

[Alexey Proskuryakov]

Skipped the test in debug builds in <http://trac.webkit.org/r163951>.

[Michael Catanzaro]

Probably a dup of bug #128395?