120451 - Crashes when waking from sleep after running Mozilla

Status: VERIFIED FIXED

(Core :: JavaScript Engine, defect, P2)

Description

[Simon Fraser [no longer active]]

I've been seeing occasional crashes when waking from sleep after running Mozilla 
on a machine (OS 9), in SleepQRemove.

I note that code was added to prmjtime.c to add a SleepQ entry to sync the clock 
after sleep. However, contrary to the docs, we never call SleepQRemove to remove 
this sleep Q entry. I strongly suspect that this is the cause of the crashes.

Comment 1

[Simon Fraser [no longer active]]

Serious bug, can cause data loss in other applications. Need to fox for 0.9.8.

Comment 2

[Simon Fraser [no longer active]]

Attachment: StdLog for one such crash.

Comment 3

[Phil Schwartau]

For reference, the changes to prmjtime.c were for bug 115545,
"JS Date() no longer accurate after system wakes from sleep"

Comment 4

[Simon Fraser [no longer active]]

Attachment: Fix to call SleepQRemove from the CFM terminate routine

This should fix the crashes for CFM. What to do for Mach-O?

Comment 5

[Simon Fraser [no longer active]]

cc pink for mach-o suggestions.

Comment 6

[Robert John Churchill]

Simon: applied your changes, booted into Mac OS 9, ran Mozilla, quit it, and put
my poor Pismo to sleep (and then woke it up) several times, no crashes.  So, I'd
say your fix "WORKSFORME".    r=rjc

Comment 7

[Steve Dagley]

As an FYI, I went back and looked at the 4.x code to see where we removed the 
SleepQ entry and it was in the patch to ExitToShell so I'd agree with putting it 
in the CFM terminate routine.  No suggestions for mach-o but it'd sure be nice if 
the OS would realize your process has gone away and not try to call into is 
(maybe OS X is that smart?)

Comment 8

[Steve Dagley]

Comment on attachment 65367 [details] [diff] [review]
Fix to call SleepQRemove from the CFM terminate routine

r=sdagley

Comment 9

[Patrick C. Beard]

Comment on attachment 65367 [details] [diff] [review]
Fix to call SleepQRemove from the CFM terminate routine

sr=beard

Comment 10

[jt.marsh]

I have a TiBook running 9.2. When using some of the latest nightlies, if I run
Mozilla, quit, then shut the lid, the computer beeps and does not sleep. When I
open the lid, there is a message that says the finder quit yada yada. The
computer will nolonger go to sleep and I have to reboot. Is this the same
problem or different?

Comment 11

[Mike Pinkerton (not reading bugmail)]

asked apple re: mach-o, will post when i get the info.

Comment 12

[Steve Dagley]

since pink is busy trying to track down a regression I'll post Apple's response:

For MachO:

[BEGIN]
#pragma CALL_ON_LOAD my_mod_init_func
static void my_mod_init_func(void)
{
    /* do my init stuff */
}

#pragma CALL_ON_UNLOAD my_mod_term_func
static void my_mod_term_func(void)
{
    /* do my termination stuff */
}
[END]


That was followed up with this comment however:

Just FYI: some of the cleanup that you do on 9 (sleep Q entries, for 
example) isn't necessary on X. Also, if your app is summarily terminated 
via Force Quit from the Processes window or from the app's Dock menu, 
your termination proc isn't going to be called either.

-eric

So the answer looks to be we don't need to do anything for the mach-o

Comment 13

[Simon Fraser [no longer active]]

Taking.

Comment 14

[Simon Fraser [no longer active]]

I'll #ifdef !XP_MACOSX around the __JSInitialize and __JSTerminate so that this 
builds for Mach-O.

Comment 15

[Christopher Blizzard (:blizzard)]

a=blizzard on behalf of drivers for 0.9.8

Comment 16

[Simon Fraser [no longer active]]

Fix checked in.

Comment 17

[sairuh (rarely reading bugmail)]

*** Bug 121486 has been marked as a duplicate of this bug. ***

Comment 18

[Phil Schwartau]

Verified Fixed with trunk binary 20020125xx on Mac9.1.
Mozilla does not crash when my Mac wakes up from sleep. 

Furthermore, the fix to bug 115545 is still intact;
i.e. JS Dates are still accurate after Mac sleep...