Closed Bug 371550 Opened 18 years ago Closed 18 years ago

Crash or hang when scrolling a page while it's loading

Categories

(Camino Graveyard :: General, defect)

PowerPC
macOS
defect
Not set
critical

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: alqahira, Unassigned)

References

Details

(Keywords: crash, hang)

Attachments

(2 files)

Attached file unhelpful crash log
I've seen a few mentions of this in Talkback, and I've experienced it twice myself in the past month (one crash, one hang) but can't get it to repro at will :( Talkback reports: TB29374771x, TB29374849x, TB29347114x, TB29280414x, TBTB29653399Q (last one is mine) Might be 10.3-only.
Flags: camino1.1?
Attached file unhelpful sample
I saw this come up again several times among the libobjc/random hex crashers in beta Talkback, including a couple of comments mentioning nytimes.com. Sure enough, after about 15 minutes of clicking around on nytimes.com, I crashed (when I wasn't even really trying!). I spent about an hour running with Zombies and friends, and it would only hang, not crash (and nothing useful showed up in the output). Here's the sample, which looks terribly un-useful, too :(
Kurt, since you were so successful before with Zombies and friends, would you mind trying that again here and seeing if you have more luck than I did? > Might be 10.3-only. I've since seen Talkback on this from 10.4.8 users as well.
latest trunk (home made) 10.4.8 ppc STR 1. load this tread on the forums http://forums.mozillazine.org/viewtopic.php?t=526467 2. click on one of the screenshots, load in the same tab (that is my settings for target links) 3. Popup blocker, click 'Never allow', view image 4. back 5. notice some residue of the screenshot page at the bottom of the forum page (that is another bug). 4. quickly scroll up down (scrollbar) to see if the residue disappears, fails 5. resize window (grip at right bottom) to see if residue disappears 6. crash Thread 0 Crashed: 0 libobjc.A.dylib 0x90a3f2e8 _class_lookupMethodAndLoadCache + 112 1 libobjc.A.dylib 0x90a3f1d4 objc_msgSend + 244 2 com.apple.AppKit 0x93763fa0 -[NSWindow sendEvent:] + 6424 3 com.apple.AppKit 0x9370c8d4 -[NSApplication sendEvent:] + 4172 4 com.apple.AppKit 0x93703d10 -[NSApplication run] + 508 5 com.apple.AppKit 0x937f487c NSApplicationMain + 452 6 org.mozilla.camino 0x000032dc start + 812 7 org.mozilla.camino 0x00002fe0 start + 48 Thread 1: 0 libSystem.B.dylib 0x9001f08c select + 12 try again with the friendly zombies turned on 2007-03-03 12:05:01.317 Camino[4352] *** Selector 'mouseEntered:' sent to dealloced instance 0x6d42750 of class RolloverImageButton. Break at '-[_NSZombie methodSignatureForSelector:]' to debug. 2007-03-03 12:05:01.317 Camino[4352] *** Selector 'mouseEntered:' sent to dealloced instance 0x6d42750 of class RolloverImageButton. Break at '-[_NSZombie methodSignatureForSelector:]' to debug. 2007-03-03 12:05:01.367 Camino[4352] *** Selector 'mouseExited:' sent to dealloced instance 0x6d42750 of class RolloverImageButton. Break at '-[_NSZombie methodSignatureForSelector:]' to debug. 2007-03-03 12:05:01.367 Camino[4352] *** Selector 'mouseExited:' sent to dealloced instance 0x6d42750 of class RolloverImageButton. Break at '-[_NSZombie methodSignatureForSelector:]' to debug. 2007-03-03 12:05:03.965 Camino[4352] *** Selector 'mouseEntered:' sent to dealloced instance 0x6d42750 of class RolloverImageButton. Break at '-[_NSZombie methodSignatureForSelector:]' to debug. 2007-03-03 12:05:03.966 Camino[4352] *** Selector 'mouseEntered:' sent to dealloced instance 0x6d42750 of class RolloverImageButton. Break at '-[_NSZombie methodSignatureForSelector:]' to debug. 2007-03-03 12:05:04.011 Camino[4352] *** Selector 'mouseExited:' sent to dealloced instance 0x6d42750 of class RolloverImageButton. Break at '-[_NSZombie methodSignatureForSelector:]' to debug. 2007-03-03 12:05:04.012 Camino[4352] *** Selector 'mouseExited:' sent to dealloced instance 0x6d42750 of class RolloverImageButton. Break at '-[_NSZombie methodSignatureForSelector:]' to debug. Browser then locks up/hangs (no way to load another page) the first set of four is one I clicked on the never allow button the next ones, I was back on the forum page I'll try next with an official one.
That could well be the close button; we'll see after the latest patch for that is approved and lands.
Duh, I skipped right over the zombie output. So it *is* the close button, which gives hope that this is the same as the other bug and will be fixed by the patch there. /me crosses his fingers.
(In reply to comment #6) You mean bug 372003, right ? I'll build with that patch and give it a try.
Yep, didn't have the bug number handy.
A trunk build with the patch for bug 372003 included. 1. I could not reproduce any of the symptoms that caused crashes (comment #4 above). Good. 2. I tried hard to reproduce the problems with scrolling/holding the scrollbar/... (described by Kurt), and did not experience any crashes or hangs either. And scrolling while the page is still loading. I used a couple of slow loading sites (like http://www.politiken.dk/), and test files tailored to mimic slow network conditions.
Philippe, here's some tips to help induce the crash we're seeing on an intermittent basis with this bug. In short, it helps to stress the browser. I open up seven or eight tabs, and then set them to the Mozilla random page loader test page at http://landfill.mozilla.org/ryl/random.html . In doing this, I try to load the random test page several seconds apart in each tab, while I'm browsing on some of the sites I visit on a regular basis. If I see two tab windows loading up the same page, I'll close them. After an interval of this -- say about 10 or 15 minutes -- Camino has a tendency to blow up, requiring a force-quit. Since Phillipe seemed to have the zombie stuff under control with Mac OS 10.4.8, I figured I would hold off further testing for a copule of days, until the patch fixing bug 372003 hit the trunk (I'm not building Camino -- yet.) Hope this helps.
(In reply to comment #10) I tried hard, during an hour. 10 tabs with the random page loader test page + 3 to 5 tabs with pages that I actually read. And no problems to report. Scrolling, resizing window, closing tabs, opening new tabs.
Then, Philippe, I'm not sure what's going on. This crash only popped up for me sometime during the last week of February, and only then on an intermittent basis. And as I mentioned in comment #10, it seemed to occur when I was stressing the browser. I turned off the "restore last pages upon starting" pref in an effort to see if this did anything; I'll turn it back on and see if I notice any difference.
The patch from bug 372003 was just checked in, so please test tomorrow's build if you've been seeing this.
I haven't seen this mentioned in Talkback since bug 372003 landed, nor could philippe repro with that patch, calling this one FIXED, too. If you still see this in a recent nightly, please do re-open.
Status: NEW → RESOLVED
Closed: 18 years ago
Resolution: --- → FIXED
Flags: camino1.1? → camino1.1+
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: