693821 - Using TimeDuration/TimeStamp from static constructors should assert

Status: RESOLVED FIXED

(Core :: XPCOM, defect, P3)

Description

[Boris Zbarsky [:bzbarsky]]

The classes are possibly not initialized properly at that point.

Comment 1

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Do we want each implementation to assert what it requires, or do we want the same assertions across platforms?  The first (for the Mac one) is pretty straightforward.

Comment 2

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Attachment: add assertions to mac code

untested

Comment 3

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 566362 [details] [diff] [review]
add assertions to mac code

r=me

Comment 4

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

Prior to the patch in bug 693219 this patch yields, on a mac debug tinderbox builder:

###!!! ABORT: calling TimeDuration too early: 'gInitialized', file /builds/slave/try-osx64-dbg/build/xpcom/ds/TimeStamp_darwin.cpp, line 133
WARNING: XPCOM objects created/destroyed from static ctor/dtor: file /builds/slave/try-osx64-dbg/build/xpcom/base/nsTraceRefcntImpl.cpp, line 172
WARNING: XPCOM objects created/destroyed from static ctor/dtor: file /builds/slave/try-osx64-dbg/build/xpcom/base/nsTraceRefcntImpl.cpp, line 172
WARNING: XPCOM objects created/destroyed from static ctor/dtor: file /builds/slave/try-osx64-dbg/build/xpcom/base/nsTraceRefcntImpl.cpp, line 172
WARNING: XPCOM objects created/destroyed from static ctor/dtor: file /builds/slave/try-osx64-dbg/build/xpcom/base/nsTraceRefcntImpl.cpp, line 172
WARNING: XPCOM objects created/destroyed from static ctor/dtor: file /builds/slave/try-osx64-dbg/build/xpcom/base/nsTraceRefcntImpl.cpp, line 172
mozilla::TimeDuration::FromMilliseconds(double)+0x00000040 [/builds/slave/try-osx64-dbg/build/obj-firefox/dist/bin/XUL +0x0001C550]
__static_initialization_and_destruction_0(int, int)+0x00000036 [/builds/slave/try-osx64-dbg/build/obj-firefox/dist/bin/XUL +0x00004C9E]
global constructors keyed to _ZN13nsPresContext19PrefChangedCallbackE

Comment 5

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/5435ee09cf7b

I think that fixes this bug for now, since it will at least prevent bad things from happening again.  If it ends up doing it at the stage of somebody-landing-their-patch too often, we should revisit and make the platforms consistent in how they assert.

Comment 6

[Marco Bonardo [:mak]]

https://hg.mozilla.org/mozilla-central/rev/5435ee09cf7b