71834 - enctype form submission not working (submit button does nothing, breaks bugzilla file submission)

Status: VERIFIED WORKSFORME

(Core :: DOM: Core & HTML, defect, P1)

Description

[Doron Rosenberg (IBM)]

it's impossible to attach a file to bugzilla, the submit button is not working.
I'll try to testcase this.

Comment 1

[Doron Rosenberg (IBM)]

Attachment: testcase

Comment 2

[Doron Rosenberg (IBM)]

added testcase

Comment 3

[Vladimir Ermakov]

What build are you using? My windows 98 2001-03-12-04-Mtest works fine.

Comment 4

[Doron Rosenberg (IBM)]

i'm using a self-built linux build from today. hwaara said he sees it in win32
as well, which is why I filed. I will try to get a few more people to test this

Comment 5

[Doron Rosenberg (IBM)]

to clarify: if a ENCTYPE is set, the submit button does nothing. 

Comment 6

[Kathleen :Brade]

duplicate of #71346?

Comment 7

[Akkana Peck]

Probably not a dup; bug 71346 was marked fixed yesterday, but today's build
can't submit (in fact, this bug was filed today).  This one might be caused by
the same event blockers that are causing lots of other problems (bug 71224, bug
71836), or is it possible that it was caused by the fix to 71346?

BTW, I'm seeing it in today's release build.

Comment 8

[Doron Rosenberg (IBM)]

on bug 71346 it says that it seems to try to connect to the server, while here
the button does nothing at all, no network activity.

Comment 9

[joki (gone)]

At first glance the fix to the other blockers does not seem to have fixed this.
 I'll take a closer look.

Comment 10

[joki (gone)]

In a test case with two forms, one with enctype and one without, both submit
buttons fire click events and both forms fire submit events.  But as noted in
the other comments, the enctype form has no network activity.

Comment 11

[joki (gone)]

Well I'm seeing nsFormFrame::OnSubmit get called in both cases.  Either the test
case is working and seeming to do nothing for me since it doesn't actually have
a URL to submit to or its broken somewhere inside of nsFormFrame::OnSubmit.  Can
someone with the fixes for the other event related blockers today from
/mozilla/content take a look and see which it is?

Comment 12

[Doron Rosenberg (IBM)]

in ns4.76 linux, the testcase i attached loads a bugzilla page (but with an
error, as the bugid is missing).

Comment 13

[Darin Fisher]

So, if NS4.7 is having the same behavior... could this be a server problem?

Comment 14

[rods (gone)]

eric better take a look

Comment 15

[Eric Pollmann]

The failure occurs here - we are trying to open a stream for reading that is
already open for writing.  The stream in question is the temp file that we wrote
the form upload data to:

nsFileSpecImpl::OpenStreamForReading(nsFileSpecImpl * const 0x05ae5530) line 446
nsFileSpecImpl::GetInputStream(nsFileSpecImpl * const 0x05ae5530, nsIInputStream
* * 0x0012ec90) line 495 + 15 bytes
nsFormFrame::OnSubmit(nsFormFrame * const 0x01372dec, nsIPresContext *
0x047690a0, nsIFrame * 0x01372e44) line 927 + 19 bytes
nsFormControlHelper::DoManualSubmitOrReset(nsIPresContext * 0x047690a0,
nsIPresShell * 0x04774920, nsIFrame * 0x01372da0, nsIFrame * 0x01372e44, int 1,
int 0) line 1010
nsHTMLButtonControlFrame::MouseClicked(nsIPresContext * 0x047690a0) line 345 +
29 bytes
nsHTMLInputElement::HandleDOMEvent(nsHTMLInputElement * const 0x050e3910,
nsIPresContext * 0x047690a0, nsEvent * 0x0012f3f8, nsIDOMEvent * * 0x00000000,
unsigned int 1, nsEventStatus * 0x0012f778) line 1308

Comment 16

[Eric Pollmann]

This was broken by:

mozilla/ xpcom/ io/ nsFileStream.cpp

1.34 naving%netscape.com Mar 12 14:46   #69862, r=bienvenu sr=scc. Close the
stream in the destructor.

Because:
//----------------------------------------------------------------------------------------
NS_IMETHODIMP nsFileSpecImpl::Write(const char * data, PRInt32 requestedCount,
PRInt32 *_retval)
//----------------------------------------------------------------------------------------
{
	TEST_OUT_PTR(_retval)
	//if (!mOutputStream)
	//	return NS_ERROR_NULL_POINTER;
	if (!mOutputStream) {
		nsresult rv = OpenStreamForWriting();
		if (NS_FAILED(rv))
			return rv;
	}
	nsOutputFileStream s(mOutputStream);
	*_retval = s.write(data, requestedCount);
	return s.error();
}

When the temp local variable s goes out of scope, the destructor is called, and
the file is closed - all further writes to this file will fail.

I'll back out the change to remove the blocker (since it doesn't apparently
directly fix any critical bugs and is not 0.8.1 anyhow).  CC'ing naving

Comment 17

[Eric Pollmann]

Attachment: patch (uploaded with fixed build)

Comment 18

[Eric Pollmann]

Checked in.  To verify, upload an attachment to this bug report.  :)

Comment 19

[Doron Rosenberg (IBM)]

note that blizzard filed 71962 about form submission not fully uploading

Comment 20

[Navin Gupta]

Eric, Is another fix possible rather than backing out my checkins. #69862 
fixes a couple of mail bugs #68875 and #66795

Comment 21

[Eric Pollmann]

Probably, but file upload form submission is not the only thing that uses
nsFileSpec.  Other users:

PICS
Wallet
Silent Download
Editor
PSM Glue
Browser Persist
Plugins
Netlib
Preferences
Profile Manager
XPInstall
HTML Parser

Any of these users that use the FileSpec handle to write more than one time to a
file will also be broken.  My guess is that you will either need to fix the
FileSpec Write implementation along with your fix for bug 69862, or replace all
of these users with nsIFile and the other modern file apis.

Comment 22

[Vladimir Ermakov]

When I click Submit on Build 2001-03-29-04 On windows 98 the status bar shows
resolving host bugzilla.mozilla.org... and nothing happens. Reopening

Comment 23

[Eric Pollmann]

Looking at this...

Comment 24

[Eric Pollmann]

This works in my 032822 Win NT debug build, and the checkins between 10PM and
4AM yesterday look pretty innocent.  Have you been able to reproduce this on any
platform besides Win98?  Could it be a DNS problem?

Comment 25

[James Green]

Doron's testcase works fine here on linux opt cvs tip build from the tree
closure today.

Comment 26

[Eric Pollmann]

I just pulled a commercial optimized Win NT build from this morning 032909, and
it works fine.  Marking this WORKSFORME unless someone can confirm that this bug
is reproducible...

Comment 27

[Vladimir Ermakov]

Attachment: test attachment on Linux 2001-03-29-08-Mtrunk

Comment 28

[Vladimir Ermakov]

I've attached a file on linux without problems, but on windows 98
build 2001-03-29-04-Mtrunk I cannot submit... it times out

Comment 29

[Vladimir Ermakov]

Attachment: test on 2001-04-10-04-trunk Windows 98

Comment 30

[Vladimir Ermakov]

verifying