Created attachment 503410 [details] reftest log http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1294892448.1294893752.26076.gz Rev3 WINNT 5.1 mozilla-central opt test reftest on 2011/01/12 20:20:48 s: talos-r3-xp-017 REFTEST TEST-START | http://localhost:4444/1294892631155/27/bugs/621918-2.svg REFTEST TEST-START | http://localhost:4444/1294892631155/27/bugs/621918-2-ref.svg REFTEST TEST-UNEXPECTED-FAIL | file:///C:/talos-slave/test/build/reftest/tests/layout/reftests/bugs/621918-2.svg | image comparison (==) ... REFTEST number of differing pixels: 776 REFTEST INFO | Saved log: START http://localhost:4444/1294892631155/27/bugs/621918-2.svg REFTEST INFO | Saved log: OnDocumentLoad triggering AfterOnLoadScripts REFTEST INFO | Saved log: DoDrawWindow 0,0,800,1000 REFTEST INFO | Saved log: RecordResult fired REFTEST INFO | Saved log: START http://localhost:4444/1294892631155/27/bugs/621918-2-ref.svg REFTEST INFO | Saved log: OnDocumentLoad triggering AfterOnLoadScripts REFTEST INFO | Saved log: DoDrawWindow 0,0,800,1000 REFTEST INFO | Saved log: RecordResult fired
In the reftest log, it looks like the reference case is entirely blank.
(In reply to comment #1) > In the reftest log, it looks like the reference case is entirely blank. Yeah. I have no idea why. The SVG file seems pretty simple to me, and the reference case isn't even using the code I touched in bug 621918.
Ok -- bad news & good news on the test in this bug. BAD NEWS: most of the time this testcase & reference case both render as a completely blank document, because they don't load the filter's resource doc in time, as suggested in bug 625984. (I verified this last night with a try push that added some reftest.list lines like "== 621918-2.svg about:blank" and "== 621918-2-ref.svg about:blank" -- those lines "passed" on all platforms. :-/) (So that means that in the rare occasions where the testcase "sporadically fails", it's actually just that one of the files gets lucky & loads the filter in time to render something non-blank, and the other file doesn't get as lucky.) GOOD NEWS: if we apply the <use xlink:href=filters.svg#empty> hack (noted in bug 625984 comment 8), then the testcase & reference case *stop* rendering to about:blank, and they match* each other. Verified this on try-server. *...except on Windows where they have a 1px difference on one of the characters -- but I'm assuming that's a known issue, since the other reftest for the same original bug (621918-1.svg) is flagged as "random-if(winWidget)" BOTTOM LINE: we can fix both this bug and bug 625984 with the following change (to add the filters.svg#empty hack): http://hg.mozilla.org/try/rev/d15ed75e7ae8 along with a random-if(winWidget) annotation.
Pushed the fix described at the end of Comment 13: http://hg.mozilla.org/mozilla-central/rev/90ee185d052e I think we can close this, but I'm leaving it open for now in case any currently-running test cycles end up hitting this bug and need to be tplbot-starred.
Status: NEW → RESOLVED
Last Resolved: 8 years ago
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.