80857 - Trunk & M09 crash [@ gc_find_flags()]

Status: VERIFIED FIXED

(Core Graveyard :: Installer: XPInstall Engine, defect)

Description

[Jan Carpenter]

This bug is a topcrasher for the trunk and for M09.

Here are some URLs & Comments that might help repro this crash:

     (30372757) Comments: I was downloading themes. The Native Windows theme had
justinstalled and then it crashed. I was also downloading another theme.
     (30369528) URL: fabula.eu.org
     (30369528) Comments: download/install fabula reader
     (30331844) Comments: pressing the back button through a few windows. I 
think mozilla
was installing a java class in the background
     (30307097) URL: x.themes.org 
     (30307097) Comments: I was trying to install a new theme / chrome. This is 
my first
time testing on .9 so that may be the problem but it worked fine on .8.1
     (30286253) URL: x.themes.org
     (30286253) Comments: installing a mozilla theme
     (30260610) Comments: trying to install a new theme
     (30234323) URL: http://x.themes.org
     (30234323) Comments: Trying to install a theme 'lopburi flat'. It asked me 
if I
wanted to install it(good)
     (30223111) URL: installing LCARS Chrome package from Themes.org
     (30223111) Comments: Installation completed and "completed..." message was 
shown.
Clicked ok and boom
     (30213165) Comments: I was installing a chrome and when it finished an 
error
occured.
     (30210255) Comments: trying to quick install xmlterm from xmlterm.com
     (30200341) URL: x.themes.org
     (30200341) Comments: installing Lopburi Flat chrome
     (30187293) Comments: Was installing/downloading Lopburi Flat theme.
     (30181395) Comments: installing lopburi theme off of x.themes.org
     (30178851) Comments: Installing JabberZilla 


Here is a recent stack trace:

Incident ID 30372757 
gc_find_flags() 
js_MarkGCThing() 
js_MarkGCThing() 
js_MarkGCThing() 
js_MarkGCThing() 
js_MarkGCThing() 
js_GC() 
js_ForceGC() 
js_DestroyContext() 
JS_DestroyContext() 
nsJSContext::~nsJSContext() 
nsJSContext::Release() 
nsCOMPtr_base::~nsCOMPtr_base() 
nsJSContext::CallEventHandler() 
nsJSEventListener::HandleEvent() 
nsEventListenerManager::HandleEventSubType() 
nsEventListenerManager::HandleEvent() 
nsXULElement::HandleDOMEvent() 
PresShell::HandleDOMEventWithTarget() 
nsButtonBoxFrame::MouseClicked() 
nsButtonBoxFrame::HandleEvent() 
PresShell::HandleEventInternal() 
PresShell::HandleEventWithTarget() 
nsEventStateManager::CheckForAndDispatchClick() 
nsEventStateManager::PostHandleEvent() 
PresShell::HandleEventInternal() 
PresShell::HandleEvent() 
nsView::HandleEvent() 
nsViewManager::DispatchEvent() 
HandleEvent() 
nsWidget::DispatchEvent() 
nsWidget::DispatchWindowEvent() 
nsWidget::DispatchMouseEvent() 
nsWidget::OnButtonReleaseSignal() 
nsWindow::HandleGDKEvent() 
dispatch_superwin_event() 
handle_gdk_event() 
libgdk-1.2.so.0 + 0x17cab (0x40698cab) 
libglib-1.2.so.0 + 0x104d3 (0x406c84d3) 
libglib-1.2.so.0 + 0x10b0b (0x406c8b0b) 
libglib-1.2.so.0 + 0x10bc4 (0x406c8bc4) 
nsAppShell::DispatchNativeEvent() 
nsXULWindow::ShowModal() 
nsWebShellWindow::ShowModal() 
nsContentTreeOwner::ShowAsModal() 
nsWindowWatcher::OpenWindowJS() 
nsWindowWatcher::OpenWindow() 
nsPromptService::DoDialog() 
nsPromptService::Alert() 
XPTC_InvokeByIndex() 
EventHandler() 
PL_HandleEvent() 
PL_ProcessPendingEvents() 
nsEventQueueImpl::ProcessPendingEvents() 
event_processor_callback() 
our_gdk_io_invoke() 
libglib-1.2.so.0 + 0xec9a (0x406c6c9a) 
libglib-1.2.so.0 + 0x104d3 (0x406c84d3) 
libglib-1.2.so.0 + 0x10b0b (0x406c8b0b) 
libglib-1.2.so.0 + 0x10cc1 (0x406c8cc1) 
libgtk-1.2.so.0 + 0x91f0b (0x405e7f0b) 
nsAppShell::Run() 
nsAppShellService::Run() 
main1() 
main() 
libc.so.6 + 0x18213 (0x401ea213) 


(See Bug 53123 for similar crash.)

Comment 1

[Jan Carpenter]

Adding crash, topcrash keywords

Comment 2

[Daniel Veditz [:dveditz]]

This is probably the same thing as bug 78428, but let's leave both open until 
that's proved (the other one doesn't have a stack trace).

Comment 3

[Brendan Eich [:brendan]]

Is it possible we're just blowing the stack here?

/be

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Attachment: relevant excerpt from full list of talkback stacks

Comment 5

[Brendan Eich [:brendan]]

Thanks to dbaron for sucking down all that crash data and extracting the
stacks.  Some are deep-ish, but many are shallow, so forget that stack overflow
idea.

David says the crash data clusters around 2001-05-05/07/08.  Is the crash no
longer happening?  Engineers can't seem to reproduce.  Can we confirm the kill
if it's dead?  Cc'ing Asa.

/be

Comment 6

[dbragg]

Hmm. Downloaded mulitple themes from x.themes.org as well as several of QA's
test cases.  No crashing with recent bits.  (5/18/2001 was my last pull)
dveditz' suggestion of a possible connection with 78428 may be correct.  It
seems that just downloading large files (like these themes) and some QA tests
works fine (this bug) but if the user was surfing around while downloading then
we've got 78428.  I just posted a stack trace to 78428.  Unless someone can
reproduce this bug without surfing around then I would say this bug is no longer
happening.

Comment 7

[chris hofmann]

only entry in the database for the 10 day rolling period is from may 16...

might be fixed or we may need to wait on the next milestone release to confirm

Trunkdatasample.all:
Netscape6.50,
build 2001051623,
30577913,
crash date 2001-05-17,
Windows NT  5.0 build 2195,193,1772,
UD510CDBA-2E9C-11D5-AB810090-27C21B29,,
gc_find_flags 13d496d3,
d:\builds\seamonkey\mozilla\js\src\jsgc.c,212,,

Comment 8

[chris hofmann]

wild guess...

...wondering if the domconnect landing on May 8 10:19, and the 
subsequent clean up bugs in the few days that followed might 
have made this bug go away or change in nature...  right before
we head into the JS_GC code we are executing files that jst touched
on the dom connect landing..

Comment 9

[Brendan Eich [:brendan]]

If this isn't happening any more, can we mark it WORKSFORME and get on with our
lives?  jst, jband: you guys know about any s3kr3t fixes that went in after the
xpcdom landing?

/be

Comment 10

[John Bandhauer]

I admit to not knowing what 's3kr3t' is.

Comment 11

[Syd Logan]

Do we no longer need the code review that I implied? Talking with installer team
members and rginda, it was implied to me that installer may be interacting with
JS in ways that are less than desirable. If that is not the case, then perhaps
we should mark worksforme. Otherwise, I am concerned that we might see whack the
gopher type bugs in the future. 

Comment 12

[selmer (gone)]

PDT decided to mark this fixed so we can get some testing on it and try to verify.

Comment 13

[Jan Carpenter]

The only talkback reports I find for this from today's builds are for linux.  A 
few of them refer to "installing an outdated chrome for a build before M16" but 
a couple make the comment "getting Java2 plugin". 

Comment 14

[Jimmy Lee]

Reopening bug...
Ccing Patty.

Build: 2001-05-25-09-trunk(WIN), 2001-05-25-08-trunk(MAC)

1. Launch Netscape
2. Open http://x.themes.org and click on mozilla chromes
3. Install a couple of themes (switch or not)
4. Theme does not actually change
5. Exit browser
6. Launch Netscape
7. If Profile Manager appears, choose the same one and launch

RESULT:
Crash.  Linux does not crash.  But switching to the new theme does nothing.

EXPECTED RESULT:
No Crash

TalkBack Incident ID 30919527

Stack Trace
gklayout.dll + 0x4262e (0x6034262e) 
gklayout.dll + 0x47c79 (0x60347c79) 
gkcontent.dll + 0x1b9ec (0x019fb9ec) 
gklayout.dll + 0x1c851 (0x6031c851) 
gkcontent.dll + 0x4eaac (0x01a2eaac) 
gkcontent.dll + 0x4de13 (0x01a2de13) 
gkparser.dll + 0xac23 (0x603fac23) 
gkparser.dll + 0xaec9 (0x603faec9) 
gkparser.dll + 0x8e45 (0x603f8e45) 
gkparser.dll + 0x975c (0x603f975c) 
gkparser.dll + 0x8767 (0x603f8767) 
gkparser.dll + 0x81eb (0x603f81eb) 
gkparser.dll + 0x15293 (0x60405293) 
gkparser.dll + 0x1512e (0x6040512e) 
gkparser.dll + 0x15895 (0x60405895) 
urildr.dll + 0x49e5 (0x60b049e5) 
necko.dll + 0x8089 (0x60788089) 
necko.dll + 0xc63d (0x6078c63d) 
xpcom.dll + 0x2a4dc (0x60eba4dc) 
xpcom.dll + 0x2a44a (0x60eba44a) 
xpcom.dll + 0x2a6a9 (0x60eba6a9) 

Comment 15

[Syd Logan]

Jimmy, theme switching was broken in netscape commercial. I'm assuming that is 
where you were testing this.

A fix for that bug appears to have been landed today. Marking this bug fixed so 
we can get some retesting.

Comment 16

[Jimmy Lee]

Build: 2001-05-25-09-trunk(WIN), 2001-05-25-08-trunk(MAC)

Downloading themes does not crash.  The problem was with relaunching.  Theme 
switching is a themes issue and not an install engine issue.

Marking Verified!