Since the latest places landing, test_redirectsMode.js has been orange on Thunderbird Linux (both 32 and 64 bit builds): TESTING: includeHidden(false), redirectsMode(0), maxResults(5), sortingMode(0). *** PLACES TESTS: Comparing Array to Results TEST-UNEXPECTED-FAIL | /buildbot/comm-central-trunk-linux-opt-unittest-xpcshell/build/xpcshell/tests/toolkit/components/places/tests/queries/head_queries.js | 5 == 3 - See following stack: JS frame :: /buildbot/comm-central-trunk-linux-opt-unittest-xpcshell/build/xpcshell/head.js :: do_throw :: line 439 JS frame :: /buildbot/comm-central-trunk-linux-opt-unittest-xpcshell/build/xpcshell/head.js :: do_check_eq :: line 491 JS frame :: /buildbot/comm-central-trunk-linux-opt-unittest-xpcshell/build/xpcshell/tests/toolkit/components/places/tests/queries/head_queries.js :: compareArrayToResult :: line 345 JS frame :: /buildbot/comm-central-trunk-linux-opt-unittest-xpcshell/build/xpcshell/tests/toolkit/components/places/tests/queries/test_redirectsMode.js :: check_results_callback :: line 125 JS frame :: /buildbot/comm-central-trunk-linux-opt-unittest-xpcshell/build/xpcshell/tests/toolkit/components/places/tests/queries/test_redirectsMode.js :: cartProd :: line 179 JS frame :: /buildbot/comm-central-trunk-linux-opt-unittest-xpcshell/build/xpcshell/tests/toolkit/components/places/tests/queries/test_redirectsMode.js :: run_test :: line 307 JS frame :: /buildbot/comm-central-trunk-linux-opt-unittest-xpcshell/build/xpcshell/head.js :: _execute_test :: line 322 JS frame :: -e :: <TOP_LEVEL> :: line 1 I'm currently investigating, but ideas would be appreciated. From what I can tell, SeaMonkey doesn't have this issue. So if its something to do with the build-configuration set ups, then it would be most likely an IPC versus non-IPC issue.
Local non-libxul build couldn't reproduce. Therefore trying a clobber on the boxes and a libxul build locally.
hm, we had almost permaorange on windows, thus the test is disabled on Windows, but we were unable to reproduce the problem. If you find a way to make it reproduceable, I'd be pretty happy.
and if the failure sticks, we can temporarily disable it completely. but would be really awesome for us to have a working way to reproduce on a local build.
This passes locally with a debug build of libxul Thunderbird. I guess I'm going to have to try opt next, though I might see if I can grab a builder later and take a closer look.
I can just tell you that I ran this test in a loop 6000 times on Win, and was never able to reproduce a failure :(
I've just reproduced this by downloading the packaged test that's been built on the builders and running locally in my linux vm. On my VM it seems random with about 50% failure rate. Additionally the tree has just had its first green run (out of about 10). So very frequent orange, probably timing related at a guess. FTR the steps are: # wget http://ftp.mozilla.org/pub/mozilla.org/thunderbird/tinderbox-builds/comm-central-linux/1292518869/thunderbird-3.3a2pre.en-US.linux-i686.tar.bz2 http://ftp.mozilla.org/pub/mozilla.org/thunderbird/tinderbox-builds/comm-central-linux/1292518869/thunderbird-3.3a2pre.en-US.linux-i686.tests.zip # tar xfj thunderbird-3.3a2pre.en-US.linux-i686.tar.bz2 # unzip -o thunderbird-3.3a2pre.en-US.linux-i686.tests.zip bin* certs* xpcshell* # mkdir thunderbird/plugins # cp bin/xpcshell thunderbird # cp -R bin/components/* thunderbird/components/ # cp -R bin/plugins/* thunderbird/plugins/ # python -u xpcshell/runxpcshelltests.py thunderbird/xpcshell --test-path=test_redirectsMode.js xpcshell/tests/toolkit/components/places/tests/queries/
From what I've gleaned so far. This is the section of the test we're having trouble with, but when it actually passes: TESTING: includeHidden(false), redirectsMode(0), maxResults(5), sortingMode(0). *** PLACES TESTS: Comparing Array to Results TEST-PASS | /home/cltbld/tests/xpcshell/tests/toolkit/components/places/tests/queries/head_queries.js | [compareArrayToResult : 387] 5 == 5 *** PLACES TESTS: testing testData[http://1.example.com/] vs result[http://1.example.com/] *** PLACES TESTS: testing testData[http://2.example.com/] vs result[http://2.example.com/] *** PLACES TESTS: testing testData[http://3.example.com/] vs result[http://3.example.com/] *** PLACES TESTS: testing testData[http://1.redirect.temp.example.com/] vs result[http://1.redirect.temp.example.com/] *** PLACES TESTS: testing testData[http://2.redirect.temp.example.com/] vs result[http://2.redirect.temp.example.com/] *** PLACES TESTS: Comparing Array to Results passes When the tests fails, those two redirects aren't there.
yes, so I tried with David Dahl to reproduce this on a real linux box, and we were unable to. So sounds like this could be some sort of VM issue, maybe with timers.
This is a log of the failing mode. The interesting bits to note are: - This is the first of the 'includeHidden(false)' tests. - When it fails, the http://*redirect*/ lines show that hidden is set to 1, as per the attached log. Adding dump_table("moz_places") appears to improve the failure rate, but it still isn't perfect (probably 1 in 5 rather than 1 in 2).
please try this one.
I've done 200 runs with some with busy CPU and some without and not had a failure - it looks good to me.
Assignee: bugzilla → mak77
Comment on attachment 498196 [details] [diff] [review] patch v1.0 r=sdwilsh
Status: NEW → RESOLVED
Closed: 9 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla2.0b9
You need to log in before you can comment on or make changes to this bug.