Closed Bug 627776 Opened 11 years ago Closed 11 years ago

sporadic layout/reftests/bugs/621253-1.xhtml | image comparison (==)

Categories

(Core :: SVG, defect)

defect
Not set
normal

Tracking

()

RESOLVED FIXED

People

(Reporter: dholbert, Assigned: dholbert)

References

Details

(Keywords: intermittent-failure)

Attachments

(1 file, 1 obsolete file)

http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1295631457.1295632149.24481.gz
Rev3 MacOSX Snow Leopard 10.6.2 mozilla-central opt test reftest on 2011/01/21 09:37:37
s: talos-r3-snow-021

REFTEST TEST-UNEXPECTED-FAIL | file:///Users/cltbld/talos-slave/test/build/reftest/tests/layout/reftests/bugs/621253-1.xhtml | image comparison (==)

Reftest log just had the testcase being blank.

I think this might just need the <use xlink:href="[...]#empty"> trick that we have in other filters-related reftests, to make sure the external resource document loads in time.
Assignee: nobody → dholbert
Status: NEW → ASSIGNED
OS: Linux → Mac OS X
Hardware: x86 → x86_64
OS: Mac OS X → All
Hardware: x86_64 → All
Attached patch fix: rewrite reftests (obsolete) — Splinter Review
Ok, so this reftest actually has two big problems:
 (1) it does indeed need the <use> hack, as I suspected in comment 0 (without that, it renders to blank sometimes)
 (2) Once I've added the <use> hack, the test passes both **with and without** the patch that it's supposed to be testing.  So: it's not actually an effective test for bug 621253 right now.

(When I originally wrote & ran this test, I think I must have just gotten "lucky" and hit a sporadic failure in my pre-patch reftest run, with no sporadic failure in the post-patch reftest run -- and that fooled me into thinking the test was effective.)

So it turns out that the underlying problem in bug 621253 is actually only exposed on reload or tab-switch.  So, I've reworking the reftests as follows:
 621253-helper-[external|internal]Filter.xhtml  <--- for use in iframes
 621253-1-[external|internal]Filter.html       <-- testcases
 621253-1-ref.html                 <-- reference (just a black square)

The "helper" files just have some content with a filter applied, and their wrapper-testcases just stick the helpers in an iframe with some reftest-zoom, and apply a reload.

(There are also "-2.html/-2-ref.html" variants that are identical to -1 except for a larger reftest-zoom factor.)

I've verified that the reftests 621253-[12]-externalFilter.html in this patch reliably fail when I disable bug 621253's fix, and all the tests reliably pass when I reapply the patch.

(The "internalFilter" variants pass with & without bug 621253's fix, because the bug had to do with *external* resources. I'm just including the "internalFilter" variants for theoretical completeness.)
Attachment #507982 - Flags: review?(roc)
oops - the first attachment had some unnecessary HTTP(..) annotations for the internal-filters variants. (I'd removed them locally, but I accidentally qref'd that change into a different patch.)

I've fixed that in this version.
Attachment #507986 - Flags: review?(roc)
Attachment #507982 - Attachment is obsolete: true
Attachment #507982 - Flags: review?(roc)
Landed fix v1b:
http://hg.mozilla.org/mozilla-central/rev/fcc4f82e4d3c
Status: ASSIGNED → RESOLVED
Closed: 11 years ago
Resolution: --- → FIXED
Whiteboard: [orange]
You need to log in before you can comment on or make changes to this bug.