254527 - TB09 crashes when using a symlink to a signature file [@ nsLocalFile::SetRelativeDescriptor]

Status: RESOLVED FIXED

(Thunderbird :: Message Compose Window, defect)

Description

[José Miguel Parrella Romero]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7) Gecko/20040615 Firefox/0.9
Build Identifier: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7) Gecko/20040615 Firefox/0.9

This was first experienced when opening a mailto:// link from Firefox 0.9 in
Gnome (Debian Sarge/Sid) and then reproduced later clicking New Mail in
Thunderbird 0.7.3. Compose mail opens and shows partially, then crashes. Quality
Feedback Agent opens up after crash. I imported my mail and settings (whole
profile) from Mozilla Mail, by making a symbolic link named "default.xwo" in
~/.thunderbird to my profile in ~/.mozilla (not moving the whole folder because
it's big), so I guess that's the deal.

Reproducible: Always
Steps to Reproduce:
1. Import your Mozilla 1.x profile manually into Thunderbird
2. Open a New Mail window

Actual Results:  
Mail crashed, Quality Feedback Agent opened up.

Expected Results:  
Window should have opened and let user introduce text.

Using default theme.

Comment 1

[José Miguel Parrella Romero]

Attachment: Mail entries from prefs.js file of the profile that is being used

This is a plain text "cat prefs.js | grep mail" file from the prefs.js file
which is being used in that profile.

Comment 2

[Adam Hauner]

Jose Miguel: Could you provide TalkBack Quality Agent incident ID? To get the
Talkback ID's go to your Thunderbird directory and go to components/, there
start Talkback, it'll give you a list with the IDs. Thank's a lot.

Comment 3

[José Miguel Parrella Romero]

(In reply to comment #2)
> Jose Miguel: Could you provide TalkBack Quality Agent incident ID? To get the
> Talkback ID's go to your Thunderbird directory and go to components/, there
> start Talkback, it'll give you a list with the IDs. Thank's a lot.

It's: TB494519H

Greetings.

Comment 4

[Adam Hauner]

TB494519:
nsLocalFile::SetRelativeDescriptor()
nsPrefBranch::GetComplexValue()
NS_GetPersistentFile()
nsMsgIdentity::GetSignature()
nsMsgCompose::ProcessSignature()
nsMsgCompose::BuildBodyMessageAndSignature()
nsMsgCompose::InitEditor()
XPTC_InvokeByIndex()
XPCWrappedNative::CallMethod(XPCCallContext&, XPCWrappedNative::CallMode)()
XPC_WN_CallMethod()
js_Invoke()
js_Interpret()
js_Invoke()
nsXPCWrappedJSClass::CallMethod()
nsXPCWrappedJS::CallMethod()
PrepareAndDispatch()
nsCommandManager::CommandStatusChanged()
nsComposerCommandsUpdater::UpdateOneCommand()
nsComposerCommandsUpdater::NotifyDocumentCreated()
nsEditor::NotifyDocumentListeners(nsEditor::TDocumentListenerNotification)()
nsEditor::PostCreate()
nsEditingSession::SetupEditorOnWindow()
nsEditingSession::EndDocumentLoad()
nsEditingSession::OnStateChange()
nsDocLoaderImpl::FireOnStateChange()
nsDocLoaderImpl::doStopDocumentLoad()
nsDocLoaderImpl::DocLoaderIsEmpty()
nsDocLoaderImpl::OnStopRequest()
nsLoadGroup::RemoveRequest()
nsInputStreamChannel::OnStopRequest()
nsInputStreamPump::OnStateStop()
nsInputStreamPump::OnInputStreamReady()
nsInputStreamReadyEvent::EventHandler()
PL_HandleEvent()
PL_ProcessPendingEvents()
nsEventQueueImpl::ProcessPendingEvents()
event_processor_callback()
libglib-2.0.so.0 + 0x49f5f (0x404eef5f)
libglib-2.0.so.0 + 0x24a02 (0x404c9a02)
libglib-2.0.so.0 + 0x25af8 (0x404caaf8)
libglib-2.0.so.0 + 0x25e30 (0x404cae30)
libglib-2.0.so.0 + 0x26473 (0x404cb473)
libgtk-x11-2.0.so.0 + 0x117e83 (0x401bfe83)
nsAppShell::Run()
nsAppShellService::Run()
xre_main()
main()
libc.so.6 + 0x15dc6 (0x409eddc6)

there are over 70 crashes with this signature over M17 and Thunderbird aviary
branch:
http://talkback-public.mozilla.org/talkback/fastfind.jsp?search=1&searchby=stacksig&match=contains&searchfor=nsLocalFile%3A%3ASetRelativeDescriptor&vendor=All&product=All&platform=All&buildid=&sdate=&stime=00%3A00%3A00&edate=&etime=23%3A59%3A59

Comment 5

[José Miguel Parrella Romero]

(In reply to comment #4)
> TB494519:

Found something. Setting manually
"user_pref("mail.identity.id1.attach_signature", true);" to false in prefs.js in
the profile makes Thunderbird to work OK (by now). The signature which this
setting is making reference is local and exists (and worked with Mozilla Mail).
Guess it has something to do with Firebird parsing sloppy prefs.js from Mozilla
Mail?

Comment 6

[Jay Patel [:jay]]

*** Bug 262280 has been marked as a duplicate of this bug. ***

Comment 7

[Jay Patel [:jay]]

Adding topcrash keyword and TB08 for tracking. This looks like a topcrasher for
Thunderbird 0.8.  Although users appear to be crash doing different tasks, the
stacks all look the same:
http://talkback-public.mozilla.org/talkback/fastfind.jsp?search=1&searchby=stacksig&match=contains&searchfor=nsLocalFile%3A%3ASetRelativeDescriptor&vendor=All&product=All&platform=All&buildid=&sdate=&stime=&edate=&etime=

Comment 8

[Scott MacGregor]

Jose, can you post the value of your signature file preference in prefs.js?

I want to see how we were expressing your relative signature path in prefs.js. I
think that's why we crash.

The pref would look something like:

user_pref("mail.identity.id1.sig_file", "C:\\temp\\signature.txt");

Comment 9

[Scott MacGregor]

3rd time at adjusting the summary is a charm...hopefully.

Comment 10

[timeless]

*** Bug 268779 has been marked as a duplicate of this bug. ***

Comment 11

[Scott MacGregor]

Adam, Jose, anyone who sees this crash can you post the fore mentioned signature
pref in your prefs.js file that cause this crash?

Thanks!

Comment 12

[Jay Patel [:jay]]

Updating summary with TB09 for tracking

Comment 13

[tyl2]

Scott, isn't the line
user_pref("mail.identity.id1.sig_file", "/mnt/hdb1/documentos/jose-firmas");

in the pref.js file that Jose attached to this bug?

Comment 14

[timeless]

*** Bug 274701 has been marked as a duplicate of this bug. ***

Comment 15

[jstaerk]

Please have a look at the dupe, bug 274701. 
As far as I understood simply the (wrong?) path of the sig file made my TB
crash, David helped me to create a workaround (disabling the sig in seamonkey
and manually deleting the path information in the prefs.js, see patch to the
prefs.js).

Comment 16

[timeless]

*** Bug 275058 has been marked as a duplicate of this bug. ***

Comment 17

[timeless]

*** Bug 277894 has been marked as a duplicate of this bug. ***

Comment 18

[timeless]

*** Bug 282287 has been marked as a duplicate of this bug. ***

Comment 19

[Adam Hauner]

*** Bug 286067 has been marked as a duplicate of this bug. ***

Comment 20

[timeless]

*** Bug 286067 has been marked as a duplicate of this bug. ***

Comment 21

[timeless]

*** Bug 289464 has been marked as a duplicate of this bug. ***

Comment 22

[Matthias Versen [:Matti]]

*** Bug 295984 has been marked as a duplicate of this bug. ***

Comment 23

[David :Bienvenu]

Attachment: proposed fix

proposed fix - there doesn't seem to be any contract that GetParent won't
return a null parent but an OK result, so I've added a check for that. The
windows code was returning NS_OK w/o even initializing *aParent, which I doubt
was the intent of that code.

I'm not sure what you want to do here, Darin (or if you want to be the person
to worry about this) - checking the return of GetParent instead of crashing
seems like the safest thing to do. At first glance, I don't think GetParent
should return a null parent and an NS_OK response, but maybe that is the
contract.

Comment 24

[Scott MacGregor]

cc'ing Darin given David's question in comment 23 :)

Comment 25

[David :Bienvenu]

Comment on attachment 186869 [details] [diff] [review]
proposed fix

darn, I meant to ask Darin for a review...

Comment 26

[Darin Fisher]

Comment on attachment 186869 [details] [diff] [review]
proposed fix

r+sr=darin

Comment 27

[Darin Fisher]

by the way, please fix the indentation in nsLocalFileCommon.cpp to be consistent
with surrounding code.

Comment 28

[David :Bienvenu]

Comment on attachment 186869 [details] [diff] [review]
proposed fix

I'll put in a four space indent in that file before checking in.