Magic-Mirror/gecko-dev
1
0
Fork 0
mirror of https://github.com/mozilla/gecko-dev.git synced 2024-11-08 12:37:37 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity
180,691 Commits 615 Branches 98 Tags 5.9 GiB
94660be05a
Commit Graph

61 Commits

Author SHA1 Message Date
dbaron@dbaron.org
91e460c07f Make suspectCurrent assertion not fire constantly after we Fault. b=386664 r+sr=peterv 2007-07-03 14:21:57 -07:00
dbaron@dbaron.org
212a6a5311 Change cycle collector aging constant from 10 to 0. b=381199 r+sr=peterv 2007-06-27 13:41:26 -07:00
dbaron@dbaron.org
088d801720 Fix gcc warning about casting string literal to char*, DEBUG-only. 2007-06-27 13:40:52 -07:00
dbaron@dbaron.org
0ecbf11058 Keep the purple bit in sync with the purple buffer when suspect and forget calls are dropped on the floor. b=385548 r=graydon 2007-06-27 10:51:48 -07:00
dbaron@dbaron.org
2211b60f82 Back out 385548 to try to fix test orange. 2007-06-26 22:26:38 -07:00
dbaron@dbaron.org
247f9667fd Don't set the purple bit when suspect calls are dropped on the floor. b=385548 r=graydon 2007-06-26 21:47:40 -07:00
dbaron@dbaron.org
7c5b1319a0 Print the warning message for failure to suspect or unlink only for the objects that are the roots of what would have been collected. b=385549 r=graydon 2007-06-25 13:36:24 -07:00
dbaron@dbaron.org
a957bfec45 Back out again. 2007-06-21 17:30:41 -07:00
dbaron@dbaron.org
11208f0393 See if using mScanDelay of 1 instead of 0 fixes the leak regression. 2007-06-21 16:49:37 -07:00
dbaron@dbaron.org
071c76a324 Change cycle collector aging constant from 10 to 0. b=381199 r+sr=peterv 2007-06-21 16:02:53 -07:00
graydon@mozilla.com
31b7fc133f Bug 384942, dump graph on cycle collection fault. 
r=dbaron
 2007-06-18 18:29:10 -07:00
dbaron@dbaron.org
37f48fed5f Back out bug 381199. 2007-06-13 17:12:16 -07:00
dbaron@dbaron.org
d0fb8f5ccc Change cycle collector aging constant from 10 to 0. b=381199 r+sr=peterv 2007-06-13 16:22:19 -07:00
peterv@propagandism.org
21da11d098 Fix for bug 383651 (Re-entrancy into nsCycleCollector::Collect can result in a crash). Patch by ak.miller@auckland.ac.nz, r=graydon, sr=dbaron. 2007-06-08 03:17:21 -07:00
dbaron@dbaron.org
35b99373e1 Don't loop through all the possible shutdown collections when we suspect all native wrappers. b=383234 r+sr=peterv 2007-06-06 15:09:00 -07:00
dbaron@dbaron.org
71c86c8017 Shutdown in one cycle when we set mScanDelay to 0. b=383234 r+sr=peterv 2007-06-06 15:08:25 -07:00
dbaron@dbaron.org
387bbbccf2 strdup when allocating mName so we have useful pointers. b=383234 r+sr=peterv 2007-06-06 15:07:29 -07:00
dbaron@dbaron.org
49261f75e3 Print info about things that are suspected on the extra collections at shutdown to catch deficiencies in traversal that cause us to need the extra collections. b=383234 r+sr=peterv 2007-06-06 15:06:56 -07:00
dbaron@dbaron.org
8c714701dc When explaining why expected garbage was not collected, list the known references when some were unknown. b=383234 r+sr=peterv 2007-06-06 15:06:00 -07:00
dbaron@dbaron.org
d1fbeecd5c Make DEBUG_CC compile again. b=383234 r+sr=peter 2007-06-06 15:04:26 -07:00
peterv@propagandism.org
5acbaf5478 Fix for bug 368774 (Make cycle collector work with refcounted non-XPCOM objects). r=dbaron, sr=sicking. 2007-05-24 07:10:02 -07:00
jonas@sicking.cc
233450d9b8 Relanding bug 378987. Don't try to collect windows, documents or elements of actively viewed pages. r=bz/jst sr=jst 2007-05-10 13:21:12 -07:00
jst@mozilla.org
eedb90c553 Backing out Jonas' last changes due to tinderbox orange, per Jonas' request. 2007-05-09 17:42:33 -07:00
jonas@sicking.cc
de64d03178 Relanding bug 378987. Don't try to collect windows, documents or elements of actively viewed pages. r=bz/jst sr=jst 2007-05-09 16:33:28 -07:00
jonas@sicking.cc
594d259432 back out last patch since leaks went through the roof 2007-05-07 18:03:08 -07:00
jonas@sicking.cc
c2c495c2ef Don't attempt to cycle collect documents, windows and elements that are currently being viewed. r=bz, sr=jst b=378987 2007-05-07 16:45:25 -07:00
dbaron@dbaron.org
ca87aeb5ac Remove extra class names to address Neil's comments. b=378514 2007-05-03 16:07:26 -07:00
dbaron@dbaron.org
24d770a6a9 Record results of cycle collection traversal so we only have to call the traversal APIs once per collection. b=378514 sr=peterv r=graydon 2007-04-25 14:12:11 -07:00
peterv@propagandism.org
f66cd8c4f5 Fix for bug 376854 (CycleCollector depends on uninitialized values (ageSelectionCallback and nsPurpleBuffer::BumpGeneration)). r/sr=dbaron. 2007-04-25 11:33:55 -07:00
dbaron@dbaron.org
be3b7ec321 Followup fixes to making cycle collector use more efficient hashtables: shrink size of table and fix warnings on 64-bit machines. b=377606 sr=peterv r=graydon 2007-04-23 16:34:33 -07:00
peterv@propagandism.org
a3211baf40 Fix for bug 372110 (Make cycle-collection debugging features optional at compile time). r=graydon, sr=jst. 2007-04-21 10:43:19 -07:00
peterv@propagandism.org
533f945b5a Reorder members to make struct smaller and try to lower MH down again. 2007-04-20 01:51:31 -07:00
peterv@propagandism.org
9c898bc846 Fix for bug 377606 (Switch cycle collector to more efficient hashtables). Patch by graydon, r=peterv, sr=dbaron. 2007-04-20 01:01:01 -07:00
peterv@propagandism.org
e6f2b66bd5 Fix for bug 373693 (Cycle collector makes AddRef about 6x slower). r=graydon, sr=dbaron. 2007-04-18 03:10:35 -07:00
peterv@propagandism.org
12431cf555 Fix for bug 375075 (Stop leaking the cycle collector on shutdown). r=bsmedberg, sr=dbaron. 2007-03-27 02:49:06 -07:00
peterv@propagandism.org
826c6a72f8 Backing out change for debugging tinderbox orange. 2007-03-23 09:10:21 -07:00
peterv@propagandism.org
0366969393 Temporarily enable cycle collection timing for debug builds to figure out tinderbox orange. Will back out after qm-xserve01 tinderbox has cycled. 2007-03-23 08:11:53 -07:00
jst%mozilla.org
cd1c3c94d3 Fixing bug 374239. Add debugging code that will print time spent in various parts of nsCycleCollector::Collect(). r+sr=peterv@propagandism.org 2007-03-19 23:21:31 +00:00
peterv%propagandism.org
5a1ed9c668 Fix for bug 374096 (Cycle collector doesn't collect all the cycles it could). r=graydon, sr=dbaron. 2007-03-16 12:52:47 +00:00
graydon%mozilla.com
203839a8b5 Bug 368869, previous commit had a typo: s/bool/PRBool/, s/false/PR_FALSE/. 2007-03-15 23:47:01 +00:00
graydon%mozilla.com
47299d3e29 Bug 368869 - Make cycle collection suspect all native wrapper roots 
r=brendan, sr=jst
 2007-03-15 23:05:03 +00:00
joshmoz%gmail.com
f25eb0ef7b XPCOM cycle collector is initialized statically, don't do that. b=369336 r=graydon sr=vlad 2007-03-14 06:15:10 +00:00
dbaron%dbaron.org
e19b9e8e92 Add debugging code to cycle collector to give information about objects it has been told ought to be freed. b=372766 r=graydon sr=peterv 2007-03-09 23:00:14 +00:00
peterv%propagandism.org
d31627390e Take care of reviewer comment (bug 373219). 2007-03-09 13:49:45 +00:00
peterv%propagandism.org
a267e57680 Fix for bug 373219 (ASSERTION: Fault in cycle collector: null XPCOM pointer returned (ptr: 0)). r/sr=dbaron. 2007-03-09 13:14:06 +00:00
bzbarsky%mit.edu
8cfe770629 Relanding bug 368523 now that we've established it's the reason for the Txul change. 2007-03-06 19:25:10 +00:00
bzbarsky%mit.edu
ffa520357a Back out bug 368523 to test performance impact 2007-03-06 18:23:15 +00:00
dbaron%dbaron.org
3d3dda25c0 Make nsCycleCollector::Fault do something in debug builds. b=372320 r=graydon sr=peterv 2007-03-05 21:07:51 +00:00
bzbarsky%mit.edu
5ace0b6f2a Don't hook malloc by default (still hook it if asked to via the env vars), and 
when it's hooked ignore free(0) calls.  Bug 368523, r=graydon, sr=dbaron
 2007-03-02 01:08:03 +00:00
peterv%propagandism.org
0e04a8b0f2 Fix for bug 368549 (Cycle collector doesn't handle tearoffs correctly). r=graydon, sr=jst. 2007-02-18 14:38:04 +00:00