560717 - Linux floating point exceptions running make check

Status: VERIFIED FIXED

(MailNews Core :: Build Config, defect)

Description

[Mark Banner (:standard8)]

The comm-central Linux check boxes are currently failing make check due to floating point exceptions:

TEST-PASS | TimeDuration <= operator works
TEST-PASS | TimeDuration <= operator works
TEST-PASS | TimeDuration >= operator works
TEST-PASS | TimeDuration >= operator works
TEST-PASS | TimeDuration >= operator works
TEST-PASS | Default TimeStamp value null
TEST-PASS | TimeStamp time value non-null
TEST-PASS | TimeStamp zero-length duration
TEST-PASS | TimeStamp > comparison
TEST-PASS | TimeStamp > comparison
TEST-PASS | TimeStamp < comparison
TEST-PASS | TimeStamp < comparison
TEST-PASS | TimeStamp <= comparison
TEST-PASS | TimeStamp <= comparison
TEST-PASS | TimeStamp <= comparison
TEST-PASS | TimeStamp >= comparison
TEST-PASS | TimeStamp >= comparison
TEST-PASS | TimeStamp >= comparison
TEST-PASS | TimeStamp difference lower bound
TEST-PASS | TimeStamp difference upper bound
TEST-PASS | TimeStamp negative difference lower bound
TEST-PASS | TimeStamp negative difference upper bound
/bin/sh: line 1: 13147 Floating point exceptionXPCOM_DEBUG_BREAK=stack-and-abort ../../dist/bin/run-mozilla.sh ../../dist/bin/$f
NEXT ERROR make[3]: *** [check] Error 136

Could be fallout from bug 480735.

Firefox isn't seeing these failures, so my best guesses are that it is either something related to bug 560715 (unlikely), or it could be we're using an out-of-date compiler and need to update it for the latest bug fixes.

Comment 1

[Robert Kaiser]

AFAIK, I have updated the SeaMonkey boxes to the same compiler Firefox should be using - but then, since they switched to puppet, it's hard for me to figure out what exactly they are installing on their build machines (see bug 557704).

Comment 2

[Ginn Chen]

The change of TimeStamp.cpp in Bug 441767 should also apply to TimeStamp_posix.cpp.
Otherwise it is not initialized.

Comment 3

[Ginn Chen]

Attachment: patch

Comment 4

[Serge Gautherie (:sgautherie)]

(In reply to comment #2)
> The change of TimeStamp.cpp in Bug 441767 should also apply to
> TimeStamp_posix.cpp.

s/Bug 441767/bug 480735/, right?

Comment 5

[Ginn Chen]

right.

missed 1 changeset.

Comment 6

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Attachment: patch w/add'l fix

Didn't even see _posix, sorry!  I just landed a patch that fixes a static constructor initialization hazard, but it needs a guard in TimeStamp::Startup.  This patch adds that guard (is otherwise identical to previous).

Comment 7

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

http://hg.mozilla.org/mozilla-central/rev/1e04db0f5a30

Comment 8

[Serge Gautherie (:sgautherie)]

http://tinderbox.mozilla.org/showlog.cgi?log=SeaMonkey/1272320758.1272325985.26394.gz
Linux comm-central-trunk build on 2010/04/26 15:25:58

V.Fixed