Closed Bug 775982 Opened 7 years ago Closed 7 years ago

Intermittent Android abort during indexedDB/ipc/test_ipc.html [@ nsACString_internal::MutatePrep]

Categories

(Core :: Storage: IndexedDB, defect, critical)

ARM
Android
defect
Not set
critical

Tracking

()

RESOLVED WORKSFORME

People

(Reporter: emorley, Unassigned)

References

Details

(Keywords: crash, intermittent-failure)

Crash Data

Attachments

(1 obsolete file)

See the similar bug 775053.

Android Tegra 250 mozilla-central opt test mochitest-3 on 2012-07-20 07:30:35 PDT for push 3337f0685b75

slave: tegra-225

https://tbpl.mozilla.org/php/getParsedLog.php?id=13710220&tree=Firefox

{
3863 INFO TEST-START | /tests/dom/imptests/webapps/XMLHttpRequest/tests/submissions/Ms2ger/test_setrequestheader-invalid-arguments.htm
3864 INFO TEST-PASS | /tests/dom/imptests/webapps/XMLHttpRequest/tests/submissions/Ms2ger/test_setrequestheader-invalid-arguments.htm | Elided 53 passes or known failures.
3865 INFO TEST-END | /tests/dom/imptests/webapps/XMLHttpRequest/tests/submissions/Ms2ger/test_setrequestheader-invalid-arguments.htm | finished in 662ms
3866 INFO TEST-START | /tests/dom/indexedDB/ipc/test_ipc.html
3867 INFO TEST-PASS | /tests/dom/indexedDB/ipc/test_ipc.html | Got first iframe load event.
Automation error: Error receiving data from socket (possible reboot). cmd={'cmd': 'ps'}; err=[Errno 54] Connection reset by peer
reconnecting socket

INFO | automation.py | Application ran for: 0:19:08.723036
INFO | automation.py | Reading PID log: /tmp/tmp9VeurVpidlog
getting files in '/mnt/sdcard/tests/profile/minidumps/'
Downloading symbols from: http://ftp.mozilla.org/pub/mozilla.org/mobile/tinderbox-builds/mozilla-central-android/1342791292/fennec-17.0a1.en-US.android-arm.crashreporter-symbols.zip
PROCESS-CRASH | automation.py | application crashed (minidump found)
Crash dump filename: /tmp/tmp2IcbdD/66ac5da3-e8de-422b-5b9120a0-4dcb18eb.dmp
Operating system: Linux
                  0.0.0 Linux 2.6.32.9-00002-gd8084dc-dirty #1 SMP PREEMPT Wed Feb 2 11:32:06 PST 2011 armv7l
CPU: arm
     0 CPUs

Crash reason:  SIGSEGV
Crash address: 0x0

Thread 0 (crashed)
 0  libmozalloc.so!mozalloc_abort [mozalloc_abort.cpp:3337f0685b75 : 23 + 0x4]
     r4 = 0xafd42550    r5 = 0xbebb8424    r6 = 0x00000000    r7 = 0xffffffff
     r8 = 0x43e06b25    r9 = 0x00000001   r10 = 0xbebb8038    fp = 0x4446ab17
     sp = 0xbebb8028    lr = 0xafd162f5    pc = 0x420b9704
    Found by: given as instruction pointer in context
 1  libc.so + 0x162f4
     r4 = 0xafd42550    r5 = 0xbebb8424    r6 = 0x00000000    r7 = 0xffffffff
     r8 = 0x43e06b25    r9 = 0x00000001   r10 = 0xbebb8038    fp = 0x4446ab17
     sp = 0xbebb8028    pc = 0xafd162f5
    Found by: call frame info
 2  libxul.so!nsACString_internal::MutatePrep [nsTSubstring.cpp:3337f0685b75 : 131 + 0x0]
     sp = 0xbebb8030    pc = 0x43e11db5
    Found by: stack scanning
 3  0x2323231f
     r4 = 0x00000000    r5 = 0x4446aad7    r6 = 0x6968435b    r7 = 0x3420646c
     r8 = 0x5d323836    sp = 0xbebb8048    pc = 0x23232320
    Found by: call frame info
}
Attachment #644356 - Flags: review?(bmo) → review+
Whiteboard: [orange] → [orange][test disabled][leave open]
Wait, correct me if I'm wrong here, but this test crashed once for running out of memory and so the decision was to disable the entire indexedDB OOP test suite on android?
As a followup to the 60 reported failures (bearing in mind only 10% on android failures get starred precisely, the rest generically so, due to the quantity and poor parsing of failures) in bug 775053, yes.
(In reply to Ed Morley [:edmorley] from comment #5)
> As a followup to the 60 reported failures (bearing in mind only 10% on
> android failures get starred precisely, the rest generically so, due to the
> quantity and poor parsing of failures) in bug 775053, yes.

Um, that's a totally different test:

dom/devicestorage/ipc/test_ipc.html vs. indexedDB/ipc/test_ipc.html
Bug 775053 caused the suite to halt before we reached this test, which is why there is only one occurrence in this bug (I believe there were others that ended up being starred generically and not posted here); trying to remember back to what was discussed on IRC, I believe we thought this test was exhibiting the same behaviour & so should be disabled as well. I guess we could re-enable for a day and see how frequent it is. 

Though I'm pretty much at the point of hiding native Android M3 as it is anyway, for its 70%+ failure rate, at which point whether this test is enabled or not will be academic, since no one will be paying attention to the suite anyway.

Welcome to the sorry state of tests on Android... :-(
Joel, Ed, please disable these tests ifdef MOZ_WIDGET_ANDROID if they're only to be disabled for the fennec-android products.  We're trying to keep running and passing for b2g.
Does B2G use testing/mochitest/android.json, I thought not?
I believe so, but that's a bug we're fixing.
Just to explain why this is a big deal: test_ipc.html is not a simple test but rather an entire OOP test suite. Disabling it means that we're entirely blind to breakage that could hose B2G. In fact, since the devicestorage test is also disabled, I'm not sure if we have *any* tests that run OOP on B2G now. Of course, if this test runs out of memory frequently and obscures other failures then we need to come up with something else, but we won't know unless it's re-enabled.
I've backed out the test disabling; sorry for the misunderstanding as to platforms affected & for us thinking it was more closely related to bug 775053 than it actually was.

Backout of comment 2 landing:
https://hg.mozilla.org/integration/mozilla-inbound/rev/1abfd50c8be6
Whiteboard: [orange][test disabled][leave open] → [orange]
Thanks! Now I just hope it doesn't go orange frequently :)
Attachment #644356 - Attachment is obsolete: true
(Native Android mochitest-3 is hidden at the moment due to high overall rate of failure, hence the lack of TBPLBot posts here)
indexedDB/ipc/test_ipc.html continues to fail on Native Android M3:

https://tbpl.mozilla.org/php/getParsedLog.php?id=13964532&tree=Firefox&full=1
https://tbpl.mozilla.org/php/getParsedLog.php?id=14003427&tree=Mozilla-Inbound&full=1

In many other Native Android M3 runs, test_ipc does not run at all, because of failures in dom/imptests (which I hope to disable soon).
Status: NEW → RESOLVED
Closed: 7 years ago
Resolution: --- → WORKSFORME
Whiteboard: [orange]
You need to log in before you can comment on or make changes to this bug.