1284930 - mozalloc_abort stderr message not shown in log for xpcshell job

Status: NEW

(Testing :: XPCShell Harness, defect, P3)

Description

[Dana Keeler (she/her) (use needinfo) [:keeler] (on leave)]

Clicking on the failed xpcshell test here:

https://treeherder.mozilla.org/#/jobs?repo=try&revision=30ad986fc2c1&selectedJob=23389449

Shows this:

PROCESS-CRASH | modules/libpref/test/unit_ipc/test_large_pref.js | application crashed [@ mozalloc_abort(char const*)]
Exceeded max bug suggestions
Return code: 1

which is not very helpful.

Comment 1

[Ed Morley [:emorley]]

Which bug should be suggested in this case?

I don't find any filed for the test name 'test_large_pref.js', which just leaves the '[@ mozalloc_abort(char const*)]', which is a generic abort message, for which showing bug suggestions isn't going to be useful.

If seems like the real issue here is that the raw log doesn't contain any output relating to the actual abort that caused the 'crash'.

ie: I'd have expected another log line between the TEST-START and the PROCESS-CRASH from a debug print statement.

16:06:36     INFO -  TEST-START | security/manager/ssl/tests/unit/test_pkcs11_token.js
16:06:37     INFO -  mozcrash Saved minidump as /builds/slave/test/build/blobber_upload_dir/9D85D608-0FCC-4D99-A532-5E8601CCD8B6.dmp
16:06:37     INFO -  mozcrash Saved app info as /builds/slave/test/build/blobber_upload_dir/9D85D608-0FCC-4D99-A532-5E8601CCD8B6.extra
16:06:37 WARNING - PROCESS-CRASH | modules/libpref/test/unit_ipc/test_large_pref.js | application crashed [@ mozalloc_abort(char const*)]

Comment 2

[Dana Keeler (she/her) (use needinfo) [:keeler] (on leave)]

Maybe it could suggest filing a bug for test_large_pref.js? Or say something like 'the information we have is too generic to narrow down to a matching intermittent bug', rather than suggesting that there are matches and then not showing them.

Comment 3

[Ed Morley [:emorley]]

Seems like there are a few issues here:

1) (a mozharness/xpcshell issue) We're not seeing the error printed to stderr here:
https://dxr.mozilla.org/mozilla-central/rev/70e05c6832e831374604ac3ce7433971368dffe0/memory/mozalloc/mozalloc_abort.cpp#25

2) The wording of "Exceeded max bug suggestions" could be tweaked to make it clearer. ("max" is 20 bugs, and the reason that is shown is that typically if there are that many bugs, the search term was actually just inappropriate/too common and is causing false positives.)

3) In this case 'mozalloc_abort(char const*)' should also probably be on the blacklist:
https://github.com/mozilla/treeherder/blob/0750d700cd7b108334e4dd9408f87200dca22380/treeherder/model/error_summary.py#L136-L165

(In reply to David Keeler [:keeler] (use needinfo?) from comment #2)
> Maybe it could suggest filing a bug for test_large_pref.js? Or say something
> like 'the information we have is too generic to narrow down to a matching
> intermittent bug', rather than suggesting that there are matches and then
> not showing them.

There is a WIP feature for this, bug 1117583.

Comment 4

[Ed Morley [:emorley]]

It's worth noting in this case, filing a bug for test_large_pref.js would likely be incorrect, since the root cause is an abort which may/may not be specific to a single test.

I'll file some Treeherder bugs for #2 and #3 in comment 3 when I get a second tomorrow, however the root cause of this issue is an xpcshell/mozharness issue (#1), so moving there :-)

Ted, any ideas why the stderr output here:
https://dxr.mozilla.org/mozilla-central/rev/70e05c6832e831374604ac3ce7433971368dffe0/memory/mozalloc/mozalloc_abort.cpp#25

...wasn't in the log:
https://treeherder.mozilla.org/logviewer.html#?job_id=23389449&repo=try#L4733

Comment 5

[(not currently active) Ted Mielczarek]

I think this might be a dupe of bug 1282522 (which has a patch).