Closed
Bug 13997
Opened 26 years ago
Closed 26 years ago
browser crashes when attaching a file in bugzilla
Categories
(Core :: DOM: Core & HTML, defect, P1)
Tracking
()
VERIFIED
FIXED
M12
People
(Reporter: pavlov, Assigned: pollmann)
References
Details
Attachments
(4 files)
the browser crashes when attaching a file in bugzilla
Updated•26 years ago
|
Assignee: karnaze → kmcclusk
Comment 1•26 years ago
|
||
Kevin, I tried attaching a file and Viewer hung up. I'm not sure who should get
this.
Updated•26 years ago
|
Assignee: kmcclusk → pollmann
Comment 2•26 years ago
|
||
The nsIInputStream passed to nsHTTPChannel::SetPostDataStream points to released
memory. Looks like a refcount problem in the multipart processing code in
nsFormFrame::OnSubmit at the line:
rv = NS_NewPostDataStream(!isURLEncoded, postBuffer, 0, &postDataStream);
Here is the stack when it crashes:
nsHTTPChannel::SetPostDataStream(nsHTTPChannel * const 0x021e0920,
nsIInputStream * 0x02204ee0) line 754 + 21 bytes
nsDocumentBindInfo::Bind(nsIURI * 0x021b7ba0, nsIStreamListener * 0x00000000,
nsIInputStream * 0x02204ee0) line 1623
nsDocLoaderImpl::LoadDocument(nsDocLoaderImpl * const 0x00bdb970, nsIURI *
0x021b7ba0, const char * 0x0127d060, nsIContentViewerContainer * 0x00bda760,
nsIInputStream * 0x02204ee0, nsISupports * 0x00000000, nsIStreamObserver *
0x00bd95c4, unsigned int 0, const unsigned int 0) line 600 + 18 bytes
nsWebShell::DoLoadURL(nsIURI * 0x021b7ba0, const char * 0x0127d060,
nsIInputStream * 0x02204ee0, unsigned int 0, const unsigned int 0) line 2086 +
63 bytes
nsWebShell::LoadURI(nsWebShell * const 0x00bda760, nsIURI * 0x021b7ba0, const
char * 0x0127d060, nsIInputStream * 0x02204ee0, int 1, unsigned int 0, const
unsigned int 0, nsISupports * 0x00000000) line 2242 + 28 bytes
nsWebShell::LoadURL(nsWebShell * const 0x00bda760, const unsigned short *
0x022004e0, const char * 0x0127d060, nsIInputStream * 0x02204ee0, int 1,
unsigned int 0, const unsigned int 0, nsISupports * 0x00000000) line 2306 + 48
bytes
nsWebShell::LoadURL(nsWebShell * const 0x00bda760, const unsigned short *
0x022004e0, nsIInputStream * 0x02204ee0, int 1, unsigned int 0, const unsigned
int 0, nsISupports * 0x00000000) line 1911
nsWebShell::HandleLinkClickEvent(nsIContent * 0x01c36610, nsLinkVerb
eLinkVerb_Replace, const unsigned short * 0x022004e0, const unsigned short *
0x0029f1a0 gCommonEmptyBuffer, nsIInputStream * 0x02204ee0) line 3003
OnLinkClickEvent::HandleEvent() line 2830
HandlePLEvent(OnLinkClickEvent * 0x02204e80) line 2843
PL_HandleEvent(PLEvent * 0x02204e80) line 509 + 10 bytes
PL_ProcessPendingEvents(PLEventQueue * 0x00b925f0) line 470 + 9 bytes
_md_EventReceiverProc(HWND__ * 0x047003d0, unsigned int 49419, unsigned int 0,
long 12133872) line 938 + 9 bytes
USER32! 77e71250()
00b925f0()
If I add a addref (postDataStream) it doesn't crash, Instead it gets caught in
an endless loop. If I try and break into the debugger while caught in the loop
I don't get a stack trace. I think it is general problem with posting multipart
files.
Eric, reassigning to you. This may be a duplicate of one of your bugs.
| Assignee | ||
Updated•26 years ago
|
Status: NEW → ASSIGNED
| Assignee | ||
Comment 3•26 years ago
|
||
This looks vaguely like bug 7973 which says something like "multipart doesn't
work", but is simpler.
The first crash on Linux is far simpler than the one you're seeing Kevin. It is
just a deref of a null pointer in the logic that generates a temp file name.
Rather than do fix that, I'm going to use nsFileSpec and
nsSpecialSystemDirectory routines that should have been used all along.
I'm guessing this will leave the craziness that you are seeing still, but I'll
go after that next.
| Assignee | ||
Updated•26 years ago
|
Severity: normal → critical
Priority: P3 → P1
Target Milestone: M12
| Assignee | ||
Comment 4•26 years ago
|
||
Updating milestone/priority
| Assignee | ||
Comment 5•26 years ago
|
||
Okay, I finished rewriting that much, but I'm still seing the crash with the
stack trace here. Still working on this...
| Assignee | ||
Comment 6•26 years ago
|
||
Got rid of the crash. Form posting now hangs. Still working. :)
| Assignee | ||
Comment 7•26 years ago
|
||
Got multipart posting to work for one test case, but still halts when posting
attachments to bugzilla. This bugfix so far has involved a rewrite of the
ProcessAsMultipart routine, changes in OnSubmit, and several bugfixes in our
HTTP/Necko library - may or may not be approved for beta, we'll see. :)
| Assignee | ||
Comment 8•26 years ago
|
||
| Assignee | ||
Comment 9•26 years ago
|
||
| Assignee | ||
Updated•26 years ago
|
Status: ASSIGNED → RESOLVED
Closed: 26 years ago
Resolution: --- → FIXED
| Assignee | ||
Comment 10•26 years ago
|
||
Just checked in a fix.
To verify, try to attach a file to this bug using apprunner. If it works, the
bug is fixed. Note that, at least on Linux, you must select a file from the
current directory. This is being worked on - see bug 15317
Comment 11•26 years ago
|
||
I've tried attaching files for a bit here... all to no avail. (Win32 only - I'll
be trying Linux soon.) I'll keep you posted...
| Assignee | ||
Comment 12•26 years ago
|
||
This must be related to the problems you're seeing with bug 8209. I'm seeing
them too, so it is most likely a new bug with multipart form posting that
cropped up with some of the recent changes I've made. I'm working on this one,
and will update these bugs with status when it's resolved.
Comment 13•26 years ago
|
||
Comment 14•26 years ago
|
||
OK, Windows NT successfully uploaded a file. Now to try Linux...
Comment 15•26 years ago
|
||
Well, my Linux box always hangs when trying to bring up a file picker. I'm
reassigning the QA contact to elig in hopes that he'll be able to verify this on
Linux. (I used the 1999110508 build to verify on Windows NT.)
Comment 16•26 years ago
|
||
Updated•26 years ago
|
Status: RESOLVED → VERIFIED
Comment 17•26 years ago
|
||
Per cpratt's request and pollman's handy Linux note above, successfully attached
a file using 1999 11/10 08:00 Linux build.
Verifying fixed.
Updated•6 years ago
|
Component: HTML: Form Submission → DOM: Core & HTML
You need to log in
before you can comment on or make changes to this bug.
Description
•