99155 - Memory leak of 244 bytes from 1 block allocated in NS_NewScriptGlobalObject

Status: VERIFIED WORKSFORME

(SeaMonkey :: MailNews: Message Display, defect, P3)

Description

[[e-mail] (gone - use [e-mail] instead)]

MLK: Memory leak of 244 bytes from 1 block allocated in
NS_NewScriptGlobalObject(nsIScriptGlobalObject * *)
        Distribution of leaked blocks
               244 bytes from 1 block of 244 bytes (0x04b2f960) 
        Allocation location
            new(UINT)+0xc        [new.cpp:23 ip=0x002da05c]
            NS_NewScriptGlobalObject(nsIScriptGlobalObject * *)+0xf7
[c:\moz_src\mozilla\dom\src\base\nsGlobalWindow.cpp:3964 ip=0x08e58477]
            nsDOMSOFactory::NewScriptGlobalObject(nsIScriptGlobalObject *
*)+0x24 [c:\moz_src\mozilla\dom\src\build\nsDOMFactory.cpp:135 ip=0x08e13134]
            nsDocShell::EnsureScriptEnvironment(void)+0x27f
[c:\moz_src\mozilla\docshell\base\nsDocShell.cpp:5944 ip=0x08bbd1df]
            nsWebShell::GetInterface(nsID const&,void * *)+0x4c6
[c:\moz_src\mozilla\docshell\base\nsWebShell.cpp:287 ip=0x08b63fa6]
            nsGetInterface::()(nsID const&,void * *)const+0x1bc
[c:\moz_src\mozilla\xpcom\base\nsIInterfaceRequestor.cpp:38 ip=0x1004cecc]
            nsCOMPtr<nsIDOMWindowInternal>::assign_from_helper(nsCOMPtr_helper
const&,nsID const&)+0x62 [.\..\..\..\dist\include\nsCOMPtr.h:971 ip=0x04755bf2]
           
nsCOMPtr<nsIDOMWindowInternal>::nsCOMPtr<nsIDOMWindowInternal>(nsCOMPtr_helper
const&)+0x77 [.\..\..\..\dist\include\nsCOMPtr.h:552 ip=0x04767147]
            nsAppShellService::GetHiddenWindowAndJSContext(nsIDOMWindowInternal
* *,JSContext * *)+0x1f6
[c:\moz_src\mozilla\xpfe\appshell\src\nsAppShellService.cpp:763 ip=0x04779db8]

Comment 1

[Johnny Stenback (:jst)]

We need more info here, if we leak a window object, we need to know when to
track down the cause. What are the steps to reproduce this leak? In general,
we're pretty good about not leaking window objects.

Comment 2

[[e-mail] (gone - use [e-mail] instead)]

Login to a POP3 account, hit NewMsg.  Type [e-mail] as the 
recipient, type test in the subject field, and test in the body.  Hit Send.   
Read message, delete message and shutdown (didn't delete trash).

Comment 3

[Johnny Stenback (:jst)]

Stephen, can you get a longer stack trace of the code where the leaked object is
created? The reason I'm asking is that it looks like the leak comes from
nsAppShellService::GetHiddenWindowAndJSContext(), but according to lxr there's
no code in mozilla that calls that method, appart from nsAppShellService itself,
and those calls don't leak the window (not directly at least).

Comment 4

[[e-mail] (gone - use [e-mail] instead)]

jst, sorry, that's all Purify gave me.  Note that Purify, on occasion has known
to be out-of-whack, or just plain non-informative.  I'm happy to mark this
WON'TFIX or INVALID.  I just went on a leak-filing spree for mail, and this
crept up in the logs.  Only information I can offer is that it might be related
to the nsMsgProgressWindow?  We have other leaks (60 bytes) there...

Comment 5

[Navin Gupta]

Stephend, I think these are way too many actions. You will need to isolate
them. 

Comment 6

[Johnny Stenback (:jst)]

Since this is mailnews related I'm handing this off to the mailnews
component/product, I'm willing ot help out with this if this turns out to be a
leak in the DOM code, but as I said, in general we don't leak window objects in
the browser so I'm guessing that there's an unbalanced refcount somewhere in the
mailnews code.

Comment 7

[[e-mail] (gone - use [e-mail] instead)]

Is the following just a bogus warning or what?

nsMsgComposeSendListener: Success on the message copy operation!
WARNING: requested removal of nonexistent window
, file c:\moz_src\mozilla\embedding\components\windowwatcher\src\nsWindowWatcher
.cpp, line 846
WEBSHELL- = 2

Comment 8

[[e-mail] (gone - use [e-mail] instead)]

Attachment: Much better stack

Comment 9

[David :Bienvenu]

Seem to be leaking hidden window. Marking nsbeta1+ but willing to downgrade if
it's really a one-time only leak.

Comment 10

[scottputterman]

If we can prove this is happening more than once, we should bring this back.

Comment 11

[[e-mail] (gone - use [e-mail] instead)]

I don't see this any longer, sorry for the bogus report back then.

Comment 12

[[e-mail] (gone - use [e-mail] instead)]

verified