[Chromium] history-only-cached-subresource-loads*.html dump wrong type of output, from wrong test

• Component: WebCore Misc.
• Status: RESOLVED WONTFIX

[Hajime Morrita]

This started to timeout timeout between r103011 and r103034 according to the builder.
But there is no doubtful change.

See Bug 65462 for timeout on Vista.

[Adrienne Walker]

See: http://code.google.com/p/chromium/issues/detail?id=107819

[Adrienne Walker]

Committed r103084: <http://trac.webkit.org/changeset/103084>

[Joshua Bell]

At some point recently these flipped from MISSING expectations to TEXT differences on Linux, so updated the platform/chromium/test_expectations.txt entry

[Peter Kasting]

This flipped to TEXT on Linux because pixel baselines were erroneously checked in.  It should have been MISSING because the test expects to be dumped as text but for some reason is generating pixel results.

I'm removing the erroneous pixel baselines and updating the expectations.  Also trying to understand why this is failing.

[Ryosuke Niwa]

Created attachment 155993 [details]
history-only-cached-subresource-loads-max-age-https-actual.png

Looking at these results, PNG image is coming from some other test. For example,
http/tests/cache/history-only-cached-subresource-loads-max-age-https-actual.txt

layer at (0,0) size 800x600
  RenderView at (0,0) size 800x600
layer at (0,0) size 800x600
  RenderBlock {HTML} at (0,0) size 800x600
    RenderBody {BODY} at (8,8) size 784x584
      RenderText {#text} at (0,0) size 776x174
        text run at (0,0) width 776: "It's possible for a successful revalidation to cause events to fire synchronously. We shouldn't crash if one of those events stops the"
        text run at (0,155) width 360: "load. See https://bugs.webkit.org/show_bug.cgi?id=72762. "
      RenderText {#text} at (0,0) size 0x0
      RenderEmbeddedObject {OBJECT} at (360,20) size 300x150

And this output appears to be coming from cancel-during-revalidation-succeeded.html.

[Peter Kasting]

The two tests here have a couple of problems.

(1) The tests have dumpAsText() in them, but on most platforms most of the time, they're instead dumping a pixel output and a render tree.
(2) The actual pixel output doesn't even come from this test, as noted in comment 5 -- it looks like the output of http/tests/cache/cancel-during-revalidation-succeeded.html .

This could be a DRT bug, or (maybe more likely?) an actual history or cache issue.

I'm not sure who knows the history/cache stuff.  Adam, I know you've done some loader work -- do you know how to triage this to a good person?

[Stephen Chenney]

Marking test failures as WontFix. Bug is still accessible and recording in TestExpectations.