startup crash in PresShell::Init

VERIFIED WORKSFORME

Status

()

Core
Layout
P3
critical
VERIFIED WORKSFORME
19 years ago
19 years ago

People

(Reporter: tor, Assigned: chris hofmann)

Tracking

Trunk
Sun
Solaris
Points:
---

Firefox Tracking Flags

(Not tracked)

Details

(Reporter)

Description

19 years ago
9/14 CVS pull on solaris/native crashes when starting up with no component.reg
and no $(HOME)/.mozilla.  It crashes before it can paint the contents of
the profile manager window.

t@1 (l@1) signal SEGV (no mapping at the fault address) in PresShell::Init at
line 517 in file "nsPresShell.cpp"
  517     mViewManager->SetViewObserver((nsIViewObserver *)this);
(dbx) where
current thread: t@1
=>[1] PresShell::Init(this = 0x1d79c8, aDocument = 0x8dde8, aPresContext =
0x156190, aViewManager = 0x15a208, aStyleSet = 0x1d4888), line 517 in
"nsPresShell.cpp"
  [2] XULDocumentImpl::CreateShell(this = 0x8dde8, aContext = 0x156190,
aViewManager = 0x15a208, aStyleSet = 0x1d4888, aInstancePtrResult = 0x154928),
line 1669 in "nsXULDocument.cpp"
  [3] DocumentViewerImpl::Init(this = 0x154900, aNativeParent = 0xc0ad0,
aDeviceContext = 0x87090, aPrefs = 0x58a68, aBounds = STRUCT, aScrolling =
nsScrollPreference_kAuto), line 397 in "nsDocumentViewer.cpp"
  [4] nsWebShell::Embed(this = 0xa7cb8, aContentViewer = 0x154900, aCommand =
0xc0bc0 "view", aExtraInfo = (nil)), line 973 in "nsWebShell.cpp"
  [5] nsDocumentBindInfo::OnStartRequest(this = 0xc4510, channel = 0x12aee0,
ctxt = (nil)), line 1896 in "nsDocLoader.cpp"
  [6] nsChannelListener::OnStartRequest(this = 0x12df38, aChannel = 0x12aee0,
aContext = (nil)), line 2234 in "nsDocLoader.cpp"
  [7] nsFileChannel::OnStartRequest(this = 0x12aee0, channel = 0x12aee0, context
= (nil)), line 821 in "nsFileChannel.cpp"
  [8] nsOnStartRequestEvent::HandleEvent(this = 0xc2260), line 206 in
"nsAsyncStreamListener.cpp"
  [9] nsStreamListenerEvent::HandlePLEvent(aEvent = 0xc2260), line 144 in
"nsAsyncStreamListener.cpp"
  [10] PL_HandleEvent(0xc2260, 0xff3dd150, 0xff3dd140, 0xff3dd128, 0xfed6c524,
0xff3c0060), at 0xff063d98
  [11] PL_ProcessPendingEvents(0x4ceb0, 0xfdb40480, 0xfe3b1ef4, 0xff3dd2c8,
0xff3dc9d0, 0xb1), at 0xff063b5c
  [12] nsEventQueueImpl::ProcessPendingEvents(this = 0x4fd68), line 118 in
"nsEventQueue.cpp"
  [13] event_processor_callback(data = 0x4fd68, source = 5, condition =
GDK_INPUT_READ), line 153 in "nsAppShell.cpp"
  [14] gdk_io_invoke(0x94358, 0x1, 0x141328, 0x1, 0xfd8d1930, 0xffbef070), at
0xfdb6fb20
  [15] g_main_dispatch(0xffbef0e0, 0xfd8d3d70, 0x0, 0xfd8d3cb0, 0xfd8d1930,
0x3), at 0xfd8aa7b4
  [16] g_main_iterate(0xffffffff, 0x1, 0xfd8a86c4, 0xfd8d3d78, 0xfd8d3ce0,
0xfd8d3d6c), at 0xfd8aaee4
  [17] g_main_run(0xc0b90, 0xc0b90, 0xfd8d1930, 0xfd8d3d70, 0x56e00,
0xfdb6faa4), at 0xfd8ab0d8
  [18] gtk_main(0xfdba40a8, 0xfdafd5e0, 0xfdae5b1c, 0xc0b90, 0x0, 0xfdba105c),
at 0xfd9fbaa8
  [19] nsAppShell::Run(this = 0x4eea0), line 379 in "nsAppShell.cpp"
  [20] nsAppShellService::Run(this = 0x4f4d8), line 453 in
"nsAppShellService.cpp"
  [21] nsProfile::LoadDefaultProfileDir(this = 0xa84e0, profstr = 0xfc7e738c
"resource:/res/profile/cpw.xul"), line 393 in "nsProfile.cpp"
  [22] nsProfile::StartupWithArgs(this = 0xa84e0, cmdLineArgs = 0x4d938), line
308 in "nsProfile.cpp"
  [23] main1(argc = 1, argv = 0xffbef64c), line 517 in "nsAppRunner.cpp"
  [24] main(argc = 1, argv = 0xffbef64c), line 578 in "nsAppRunner.cpp"

Updated

19 years ago
Assignee: troy → hyatt

Comment 1

19 years ago
Who gets XUL bugs?

Updated

19 years ago
Assignee: hyatt → waterson

Comment 2

19 years ago
reassigning to waterson

Comment 3

19 years ago
tor: you are using the _native_ Solaris compiler? if so, could you please
verify that this is a problem with egcs?
(Reporter)

Comment 4

19 years ago
I see this crash with the SUNWspro 4.2 compilers.  Using egcs-1.1 results
in the startup crash described in #13650.

Comment 5

19 years ago
I think troy fixed 13560 (or a dup of it, which I filed earlier today).
(Reporter)

Comment 6

19 years ago
Just tried a new pull with egcs, and it still crashes like 13650.
The workshop build still dies in PresShell::Init

Updated

19 years ago
Assignee: waterson → chofmann

Comment 7

19 years ago
chofmann: this is a [PP] bug on an unsupported compiler. reassign as you feel
appropriate.
(Reporter)

Comment 8

19 years ago
Ok, I think I've tracked down the problem.  mViewManager is defined as
a nsIViewManager (note the "I") in class nsPresShell.  Thus when SetViewObserver
is called on it, it attempts to run the pure virtual implementation instead of
the one probably intended in nsViewManager (no "I").  Not sure what fix would
be appropriate.

(dbx) p mViewManager->SetViewObserver
mViewManager->SetViewObserver = &nsIViewManager::SetViewObserver() at 0x0

Why the compiler didn't catch this and why it works on any platform is still
a mystery to me.
(Reporter)

Comment 9

19 years ago
Never mind the previous comment... my brain wasn't in C++ mode.  The problem
is that it's calling the pure virtual function, but not for the reason I said.
(Reporter)

Comment 10

19 years ago
Another data point: the same problem happens with the SUNWspro 5.0 compilers.

Updated

19 years ago
Severity: normal → critical
(Assignee)

Comment 11

19 years ago
tor,

If the patch end up landing in nsPresShell troy might be the
guy to review and checkin..   Is that where we think the bug is?

http://cvs-mirror.mozilla.org/webtools/bonsai/cvsblame.cgi?file=mozilla/layout/h
tml/base/src/nsPresShell.cpp
(Assignee)

Comment 12

19 years ago
tor,

If the patch end up landing in nsPresShell troy might be the
guy to review and checkin..   Is that where we think the bug is?

http://cvs-mirror.mozilla.org/webtools/bonsai/cvsblame.cgi?file=mozilla/layout/h
tml/base/src/nsPresShell.cpp
(Assignee)

Comment 13

19 years ago
tor,

If the patch end up landing in nsPresShell troy might be the
guy to review and checkin..   Is that where we think the bug is?

http://cvs-mirror.mozilla.org/webtools/bonsai/cvsblame.cgi?file=mozilla/layout/h
tml/base/src/nsPresShell.cpp
(Reporter)

Updated

19 years ago
Status: NEW → RESOLVED
Last Resolved: 19 years ago
Resolution: --- → WORKSFORME
(Reporter)

Comment 14

19 years ago
Haven't seen this bug for a couple months now.

BTW: Why was your comment appended three times?
(Assignee)

Comment 15

19 years ago
he got wonky finger... ;-)

Updated

19 years ago
Status: RESOLVED → VERIFIED

Comment 16

19 years ago
Based on the reporter's last comments, marking as verified works for me in the
Dec 01 build.
You need to log in before you can comment on or make changes to this bug.