Closed
Bug 432189
Opened 16 years ago
Closed 16 years ago
xpcshell leaks stuff when running Mochitest (autoreleased with no pool in place - just leaking)
Categories
(Core :: General, defect)
Tracking
()
RESOLVED
DUPLICATE
of bug 435853
People
(Reporter: alqahira, Unassigned)
Details
(Keywords: regression)
Attachments
(1 file)
251.25 KB,
text/plain
|
Details |
Sometime since I last ran Mochitests back in March, xpcshell has started leaking like a sinking ship on Mac OS X when running Mochitests. 10.5/Intel here, with either static/opt Minefield or shared/debug Camino. STR: 1) cd $objdir/_tests/testing/mochitest ; python runtests.py --autorun 2) Look at the terminal log and weep :( I have a 252 KB log of the leaking that I think is too big for Bugzilla, but every entry is of the general format: 2008-05-05 00:31:26.868 xpcshell[7319:613] *** _NSAutoreleaseNoPool(): Object 0x18ae21e0 of class NSCFString autoreleased with no pool in place - just leaking Stack: (0x96b4b12f 0x96a57ec2 0x96b3272d 0x2139650 0x2139e7c 0x21379a9 0x212f6a0 0x213021f 0x23ad528 0x1ab8e42 0x1abcc1a 0x11c7a3 0x11084b 0x11c88f 0x1ab46d6 0x1aaebc3 0x23ad5ca 0x23ad746 0x238229d 0x239f96a 0x23ad528 0x1ab8e42 0x1abcc1a 0x11c7a3 0x11084b 0x11b819 0xe0c8c 0x33f1 0x3475 0x457a) Mostly NSCFString and __NSFastEnumerationEnumerator objects, but also NSPathStore2, NSConcreteData, NSCFDictionary, and NSCFData objects. Gavin suggested Core:General as a good starting place for this bug; someone who knows something about this, please direct it to a good home ;)
Comment 1•16 years ago
|
||
You can get this running any single test where a jar file is requested from the HTTP server. I haven't bothered caring because it's not an obviously visible leak and doesn't directly affect any shipping product. The server wasn't written to be meticulous in avoiding circular references, so it's entirely possible some leaking, if not necessarily this particular leak, may be the server's fault; it's also test-only code. I think this might have started showing up between March 10 and April 1, based on when a report of the server "hanging" (it was blocking on an unread output buffer) first showed up (Jesse filed, it's in Core:Testing somewhere), but that's just a guess.
Comment 2•16 years ago
|
||
Is this bug about xpcshell or mochitests? The summary and some of comment 0 mentions xpcshell, but the STR mentions mochitest. In bug 397811 I enable (by default) a link to the Mac OS X address book. When I now run the address book tests on Mac I get loads of leak warnings similar to those mentioned in comment 0. This looks to me as if the xpcshell isn't setting up the NSAutoreleasePool and hence its leaking. I'm currently trying out a few things to see if I can fix it. For me, this leak warning makes debugging xpcshell tests virtually impossible on Mac due to all the warnings.
Comment 3•16 years ago
|
||
(In reply to comment #2) > Is this bug about xpcshell or mochitests? The summary and some of comment 0 > mentions xpcshell, but the STR mentions mochitest. Ok, so I just realised the title says "when running Mochitest". I still think the problem is probably the same.
Comment 4•16 years ago
|
||
(In reply to comment #0) > 10.5/Intel here, with either static/opt Minefield or shared/debug Camino. I couldn't reproduce this (on minefield static/opt, no debug?). I've filed my xpcshell unit test problem as bug 435853 and attached a patch there.
Reporter | ||
Comment 5•16 years ago
|
||
(In reply to comment #3) > (In reply to comment #2) > > Is this bug about xpcshell or mochitests? The summary and some of comment 0 > > mentions xpcshell, but the STR mentions mochitest. > > Ok, so I just realised the title says "when running Mochitest". I still think > the problem is probably the same. The xpcshell binary being the thing Mac OS X says is leaking, so yes, quite possibly.
Reporter | ||
Comment 6•16 years ago
|
||
I applied the patch in bug 435853, and it does indeed fix this for me; thanks Mark!
Status: NEW → RESOLVED
Closed: 16 years ago
Resolution: --- → DUPLICATE
You need to log in
before you can comment on or make changes to this bug.
Description
•