Closed Bug 412294 Opened 18 years ago Closed 18 years ago

crash [@ RtlpCoalesceFreeBlocks] on archive.org

Categories

(Core :: Graphics: ImageLib, defect)

x86
All
defect
Not set
critical

Tracking

()

RESOLVED DUPLICATE of bug 411874

People

(Reporter: svl-bmo, Unassigned)

References

()

Details

(Keywords: crash, regression, topcrash)

Crash Data

Seen with SeaMonkey 2008010902, WinXP - http://crash-stats.mozilla.com/report/index/803ce026-c2a8-11dc-b4b9-001a4bd43ef6?date=2008-01-14-13 - which makes me believe this isn't bug 408310. I also don't have the download manager extension installed, so this isn't bug 354624 either. (I can't find other possible dupes, despite this looking like a topcrasher.) I've reproduced with SeaMonkey 2008011403 - but the crash report for that - http://crash-stats.mozilla.com/report/index/4269bade-c2c2-11dc-9219-001a4bd43e5c?date=2008-01-14-17 - gives me a 404. The following steps reproduce this for me about 25% of the time: 1. Load http://archive.org/ (this redirects to http://www.archive.org/index.php ) 2. Search for "http://adam.blogs.bitscribe.net/2007/01/" 3. Click on the Feb 04, 2007 * link 4. (sometimes necessary before the crash will show) - scroll around in the document, and/or close the browser. I seem to have more success with reproducing when I clear my cache before step 1 - but that could be just coincidence. Top of one of the stacktraces: 0 RtlpCoalesceFreeBlocks 1 RtlpCoalesceFreeBlocks 2 msvcr80.dll@0x4c38 3 gfxImageFrame::~gfxImageFrame() d:\builds\tinderbox\seamonkeytrunk\winnt_5.2_depend\mozilla\gfx\src\shared\gfximageframe.cpp:61 4 gfxImageFrame::`vector deleting destructor'(unsigned int) 5 gfxImageFrame::Release() d:\builds\tinderbox\seamonkeytrunk\winnt_5.2_depend\mozilla\gfx\src\shared\gfximageframe.cpp:45 6 nsCOMArray_base::~nsCOMArray_base() d:\builds\tinderbox\seamonkeytrunk\winnt_5.2_depend\mozilla\objdir\xpcom\build\nscomarray.cpp:61 7 imgContainer::`vector deleting destructor'(unsigned int) 8 imgContainer::Release() d:\builds\tinderbox\seamonkeytrunk\winnt_5.2_depend\mozilla\modules\libpr0n\src\imgcontainer.cpp:73 9 nsCOMPtr_base::~nsCOMPtr_base() d:\builds\tinderbox\seamonkeytrunk\winnt_5.2_depend\mozilla\objdir\xpcom\build\nscomptr.cpp:81 10 imgRequest::~imgRequest() d:\builds\tinderbox\seamonkeytrunk\winnt_5.2_depend\mozilla\modules\libpr0n\src\imgrequest.cpp:91 11 imgRequest::`vector deleting destructor'(unsigned int) 12 imgRequest::Release() d:\builds\tinderbox\seamonkeytrunk\winnt_5.2_depend\mozilla\modules\libpr0n\src\imgrequest.cpp:77 13 imgRequestProxy::~imgRequestProxy() d:\builds\tinderbox\seamonkeytrunk\winnt_5.2_depend\mozilla\modules\libpr0n\src\imgrequestproxy.cpp:96 14 imgRequestProxy::`scalar deleting destructor'(unsigned int) 15 imgRequestProxy::Release() d:\builds\tinderbox\seamonkeytrunk\winnt_5.2_depend\mozilla\modules\libpr0n\src\imgrequestproxy.cpp:59 16 nsRefPtr<nsProxyEventObject>::assign_assuming_AddRef(nsProxyEventObject*) d:\builds\tinderbox\seamonkeytrunk\winnt_5.2_depend\mozilla\objdir\dist\include\xpcom\nsautoptr.h:944 17 nsCOMPtr_base::assign_with_AddRef(nsISupports*) d:\builds\tinderbox\seamonkeytrunk\winnt_5.2_depend\mozilla\objdir\xpcom\build\nscomptr.cpp:89 18 nsImageLoadingContent::DestroyImageLoadingContent() d:\builds\tinderbox\seamonkeytrunk\winnt_5.2_depend\mozilla\content\base\src\nsimageloadingcontent.cpp:123 19 nsHTMLImageElement::~nsHTMLImageElement() d:\builds\tinderbox\seamonkeytrunk\winnt_5.2_depend\mozilla\content\html\content\src\nshtmlimageelement.cpp:185
I have a suspicion that this issue relates to the "curator's choice" animated gifs which load in the four columns underneath the search. If I hit the "take me back" submit before at least one of these images has finished loading, the likelihood of an immediate crash seems to be about 75%. The search doesn't need to be on anything specific - just the pre-filled http:// is just as likely to crash. If no crash happens, hitting the back button, and/or shirt-reloading and/or searching again has a decent chance of crashing.
Severity: major → critical
Flags: blocking1.9?
Keywords: regression
Version: unspecified → Trunk
This is much easier to reproduce by merely browsing to http://www.archive.org/index.php and then just hit shift-reload until it crashes.
Sounds like a good testcase to run through purify. I won't be able to until Monday, perhaps sayrer can take a look.
This crashes for me on Mac as well.
OS: Windows XP → All
FWIW, I've seen no reoccurances over last few days, since checkin for bug 411874. Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9b3pre) Gecko/2008012204 Minefield/3.0b3pre ID:2008012204
(In reply to comment #8) > > Do you see this with a recent nightly? I don't know why such crash-stats links don't work :/ With Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9b3pre) Gecko/2008012204 Minefield/3.0b3pre ID:2008012204 I can't get http://www.archive.org/index.php to crash with repeated ctrl-f5'ng (comment 3)
I agree that this is now fixed. Meanwhile, since I saw bug 411874 asking for it (regardless of if this is the same bug), I went ahead and isolated the image which caused the crash. Loading this in SeaMonkey 20080114 and holding down ctrl-R crashes reliably within 2-10 seconds. It never crashes in 20080122. Unfortunately the filesize of the image is too large for bugzilla's tastes, so I've had to upload it somewhere else: http://download.yousendit.com/6D749FF32661F480 (Link will remain valid only for a week; I'll upload somewhere better tonight.)
Flags: blocking1.9? → in-testsuite?
Marking as duplicate of bug 411874 as that bug fixed this crash.
Status: NEW → RESOLVED
Closed: 18 years ago
Resolution: --- → DUPLICATE
Dolske can we get that in the image test suite?
The image is now stored at http://juima.org/otherstuff/get-item-image.php-15.gif - it'll disappear from there eventually, but not before it's in the testsuite.
I attached the GIF from comment 14 to bug 411874. [The trick is to attach it as a patch, and then change the content type back to GIF.]
Crash Signature: [@ RtlpCoalesceFreeBlocks]
You need to log in before you can comment on or make changes to this bug.