Bug 62243 – Only draw portions of tiles in contentRect

• Component: New Bugs
• Status: RESOLVED FIXED

[John Bauman]

Only draw portions of tiles in contentRect

[John Bauman]

Created attachment 96321 [details]
Patch

[John Bauman]

This patch seems to work, although I'm not certain it's logically correct. This bug is the root cause of http://code.google.com/p/chromium/issues/detail?id=84901 .

[James Robinson]

Comment on attachment 96321 [details]
Patch

Can you make a reduced test case, please?

[James Robinson]

Comment on attachment 96321 [details]
Patch

Even if this patch is correct (I haven't thought about it in detail) we'll need a regression test.

[John Bauman]

Created attachment 96346 [details]
Patch

[James Robinson]

Comment on attachment 96346 [details]
Patch

The two -expected.pngs are identical, so just put one copy of it in LayoutTests/compositing/repaint/ next to the test itself.

Also in the test body put the incantation:
if (window.layoutTestController)
  layoutTestController.dumpAsText(true);

to get it to only output the pixel results and not output any render tree dump.

[John Bauman]

Oh, I was so shocked that DRT gave the same results on two different platforms that I didn't know how to respond.

[Adrienne Walker]

Thanks for fixing this.  This looks good to me too.

Does this change not cause any other layout test failures when compositing? I've seen no end of tiny filtering changes (theoretically due to floating point math) when quad sizes have changed in the past.

[John Bauman]

Created attachment 96350 [details]
Patch

[John Bauman]

Hmm, there do seem to be some unexpected failures with this patch. I'll have to look some more and see if they're really unexpected on this computer.

[John Bauman]

Comment on attachment 96350 [details]
Patch

seems to fail some layout tests.

[John Bauman]

Created attachment 96355 [details]
Correct for offset top left corner, add rebaseline requirement

[James Robinson]

Comment on attachment 96355 [details]
Correct for offset top left corner, add rebaseline requirement

R=me, enne can you take a look and flip cq+ if you're happy?

[WebKit Review Bot]

Comment on attachment 96355 [details]
Correct for offset top left corner, add rebaseline requirement

Rejecting attachment 96355 [details] from commit-queue.

Failed to run "['./Tools/Scripts/webkit-patch', '--status-host=queues.webkit.org', '--bot-id=ec2-cq-01', '--port..." exit_code: 2

Last 500 characters of output:
t/shrink-layer.html
patching file LayoutTests/platform/chromium/test_expectations.txt
Hunk #1 FAILED at 4009.
1 out of 1 hunk FAILED -- saving rejects to file LayoutTests/platform/chromium/test_expectations.txt.rej
patching file Source/WebCore/ChangeLog
Hunk #1 succeeded at 1 with fuzz 3.
patching file Source/WebCore/platform/graphics/chromium/LayerTilerChromium.cpp

Failed to run "[u'/mnt/git/webkit-commit-queue/Tools/Scripts/svn-apply', u'--reviewer', u'James Robinson', u'--f..." exit_code: 1

Full output: http://queues.webkit.org/results/8810030

[John Bauman]

Created attachment 96395 [details]
Hopefully up to date

[James Robinson]

Comment on attachment 96395 [details]
Hopefully up to date

Will land by hand

[James Robinson]

Committed r88386: <http://trac.webkit.org/changeset/88386>