1468691 - TEST-UNEXPECTED-FAIL | comm/mailnews/compose/test/unit/test_sendMessageLater3.js (windows only)

Status: RESOLVED FIXED

(Thunderbird :: General, defect)

Description

[Jorg K (CEST = GMT+2)]

TEST-UNEXPECTED-FAIL | comm/mailnews/compose/test/unit/test_sendMessageLater3.js | xpcshell return code: 1
PROCESS-CRASH | comm/mailnews/compose/test/unit/test_sendMessageLater3.js | application crashed [@ nsIFile::HumanReadablePath()]

First seen: Sun, May 27, 2018, 00:39:48
M-C last good: c4cf8be3acc84273cf07d4c3e48a1b89c5
https://treeherder.mozilla.org/#/jobs?repo=comm-central&revision=c6dbf51f60175062c0ebf8627f523084cc0c2e69

M-C first bad: 4e9446f9e8f0a75c7ffe063f1dfb311cc9
https://treeherder.mozilla.org/#/jobs?repo=comm-central&revision=7148c6f9fce243ff25afeaee0e4799b6f16d30fd

https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=c4cf8be3acc84273cf07d4c3e48a1b89c5&tochange=4e9446f9e8f0a75c7ffe063f1dfb311cc9

Nothing in that range that looks suspicious. I can't reproduce this on Windows 10 and it also doesn't fail on Windows 10 on the tree.

Masatoshi-san, could you please take a look. Here are a few lines from the log:
https://taskcluster-artifacts.net/XQ4uYJV2TVanugE59VwRHA/0/public/logs/live_backing.log

01:12:39  WARNING -  PROCESS-CRASH | comm/mailnews/compose/test/unit/test_sendMessageLater3.js | application crashed [@ nsIFile::HumanReadablePath()]
01:12:39     INFO -  Crash dump filename: c:\users\task_1527382680\appdata\local\temp\xpc-other-rbynvd\4b0f2b17-e0f0-4120-aa22-bed47382078c.dmp
01:12:39     INFO -  Operating system: Windows NT
01:12:39     INFO -                    6.1.7601 Service Pack 1
01:12:39     INFO -  CPU: x86
01:12:39     INFO -       GenuineIntel family 6 model 63 stepping 2
01:12:39     INFO -       8 CPUs
01:12:39     INFO -  GPU: UNKNOWN
01:12:39     INFO -  Crash reason:  EXCEPTION_ACCESS_VIOLATION_READ
01:12:39     INFO -  Crash address: 0xffffffffe5e5e5e5
01:12:39     INFO -  Process uptime: 1 seconds
01:12:39     INFO -  Thread 0 (crashed)
01:12:39     INFO -   0  xul.dll!nsIFile::HumanReadablePath() [nsLocalFileWin.cpp:4e9446f9e8f0a75c7ffe063f1dfb311cc90d56cf : 3384 + 0x0]
01:12:39     INFO -      eip = 0x513b6b53   esp = 0x001ef80c   ebp = 0x001ef874   ebx = 0x0a0d6ac0
01:12:39     INFO -      esi = 0xe5e5e5e5   edi = 0x001ef88c   eax = 0x001ef814   ecx = 0x001ef814
01:12:39     INFO -      edx = 0x00000000   efl = 0x00010202
01:12:39     INFO -      Found by: given as instruction pointer in context
01:12:39     INFO -   1  xul.dll!nsMsgDBService::~nsMsgDBService() [nsMsgDatabase.cpp:4e9446f9e8f0a75c7ffe063f1dfb311cc90d56cf : 86 + 0x13]

Comment 1

[Masatoshi Kimura [:emk]]

Are you sure m_dbCache is always valid here?
https://dxr.mozilla.org/comm-central/rev/51678c5b0b3edd827c57a70f1c5baf02d44acead/mailnews/db/msgdb/src/nsMsgDatabase.cpp#86
m_dbCache is an array of raw pointers!

Comment 2

[Jorg K (CEST = GMT+2)]

Grrrrr, that's another incarnation of bug 1377692 which has just shifted :-(

Comment 3

[Jorg K (CEST = GMT+2)]

Attachment: 1468691-dont-print.patch

No point crashing here while we haven't fixed bug 1342858 :-(

Comment 4

[:aceman]

Comment on attachment 8985799 [details] [diff] [review]
1468691-dont-print.patch

Review of attachment 8985799 [details] [diff] [review]:
-----------------------------------------------------------------

Thanks.

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/comm-central/rev/0d17b436ed09
don't print names of left-open DBs on server builds to avoid crashes. r=aceman

Comment 6

[Jorg K (CEST = GMT+2)]

That didn't work, still crashing here :-(
https://treeherder.mozilla.org/#/jobs?repo=comm-central&revision=b317e384ddae3e53085a36ca8432a48666378b62&selectedJob=183589044

Comment 7

[Jorg K (CEST = GMT+2)]

I'm a bit puzzled by this, the crashing code shouldn't have been compiled in to the binary. Masatoshi-san, does OFFICIAL_BUILD not get set in C++? I can see it here:
https://searchfox.org/mozilla-central/search?q=OFFICIAL_BUILD&case=false&regexp=false&path=.h
What am I missing/doing wrong?

Comment 8

[Masatoshi Kimura [:emk]]

(In reply to Jorg K (GMT+2) from comment #7)
> I'm a bit puzzled by this, the crashing code shouldn't have been compiled in
> to the binary. Masatoshi-san, does OFFICIAL_BUILD not get set in C++? I can
> see it here:
> https://searchfox.org/mozilla-central/
> search?q=OFFICIAL_BUILD&case=false&regexp=false&path=.h
> What am I missing/doing wrong?

Note that all search results are from imported Google code. The OFFICIAL_BUILD macro is irrelevant to us.

Comment 9

[Jorg K (CEST = GMT+2)]

Thanks, I confused it with MOZILLA_OFFICIAL :-(
https://searchfox.org/comm-central/search?q=MOZILLA_OFFICIAL&case=false&regexp=false&path=.cpp

That can be used, right?

Comment 10

[Jorg K (CEST = GMT+2)]

Attachment: 1468691-follow-up.patch

Comment 11

[Jorg K (CEST = GMT+2)]

Thank you both, let's see whether we have more luck this time ;-)

Comment 12

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/comm-central/rev/d494b829635c
Follow-up: Use MOZILLA_OFFICIAL instead of OFFICIAL_BUILD. r=aceman