44536 - HTTP POST fails after http authentication

Status: VERIFIED FIXED

(Core :: Networking: HTTP, defect, P3)

Description

[Mirek Hankus]

Mozilla (2000070308) does not POST data. Changing method to GET resolves
problem, but i think it is not right way to do it.

Comment 1

[rods (gone)]

reassigning

Comment 2

[Eric Pollmann]

Reporter, please attach a URL or test case that shows this problem, or this bug 
will have to be closed out.  Thanks!

Comment 3

[rcummins]

If there was a general problem with the POST method, I think there would be a
loud outcry from the many people using Mozilla to work in Bugzilla ;)
I cannot duplicate this problem, and am marking it "worksforme", unless the
reporter can provide a specific URL or steps to reproduce...?

Comment 4

[Mirek Hankus]

Hello !! I took a closer look at all this. So here is some additional info. POST
generally works fine. Problems occure when script specified as ACTION of the form
requires authentication (in which case Mozilla posts data twice -checked logs of
the server). I will try to configure server to demonstrate this situation, but
give me some time.

Comment 5

[Eric Pollmann]

Mirek, any chance this is bug 27006?  To see, look at the source of the page 
returned by the server after the POST.  If it contains a <META> with a CHARSET 
specified, then it's probably bug 27006.  This would explain both the double 
post, and the apparent failure of posting to work at all.

If this is a duplicate of bug 27006, you can just add comments to that bug 
report.  Thanks!

Comment 6

[Mirek Hankus]

There was a META tag, but i removed it and problem occurs. Check out this URL
http://www.ce3.pl/~mhankus/form.html There is a simple form. enter any data in a
text field and press submit. You'll be prompted for username and password. Enter
user pass and password pwd   You will see output of cgi script. Line begining
with STDIN: displays all the data which appeard on the stadnard input. When form
is submited with Mozilla there is no data (netscape works fine).  One more
thing. Open this form with fresh Mozilla (i mean just started - not the one
which worked for some time).

There is one more strange behaviour. Opening URL (above), submitting from, and
then pressing reload button makes Mozilla spawn additional browsers until
systems runs out of resources. 

Comment 7

[Eric Pollmann]

I see this too.  Thanks for the test case!

Comment 8

[Eric Pollmann]

CC'ing Gagan - this is probably a networking related problem.

Comment 9

[Eric Pollmann]

Seems like a networking problem, passing over to the gurus.  This is the first 
request we make when posting the form.

POST /~mhankus/auth/a.cgi HTTP/1.1M
Referer: http://www.ce3.pl/~mhankus/form.htmlM
Host: www.ce3.plM
User-Agent: Mozilla/5.0 (Windows; U; WinNT4.0; en-US; m17) Gecko/20000709M
Accept: */*M
Accept-Language: enM
Accept-Encoding: gzip,deflate,compress,identityM
Keep-Alive: 300M
Connection: keep-aliveM
Content-type: application/x-www-form-urlencodedM
Content-Length: 8M
M
var=test

This results in a server challenge.  Then, we post this as the second request:

GET /~mhankus/auth/a.cgi HTTP/1.1
Authorization: Basic cGFzczpwd2Q=
Host: www.ce3.pl
User-Agent: Mozilla/5.0 (Windows; U; WinNT4.0; en-US; m17) Gecko/20000709
Accept: */*
Accept-Language: en
Accept-Encoding: gzip,deflate,compress,identity
Keep-Alive: 300
Connection: keep-alive

The authorization header is correct, but the request method has changed from 
POST to GET and we have lost the post data.  Nav does this correctly, and the 
second request is identical to the first with the addition of the Authorization 
header.

Comment 10

[Eric Pollmann]

Please ignore the terminal M after each line in the first request, it was a bad 
cut and paste job on my part - those are carriage returns.  ;)

Comment 11

[Gagan]

beta3

Comment 12

[Phil Peterson]

Not holding PR3 for this; marking nsbeta3-. Please nominate for rtm if we really
need to fix this before shipping Seamonkey.

Comment 13

[Gagan]

pulling in ruslan's necko bugs ->darin

Comment 14

[Gagan]

http bugs to "Networking::HTTP"

Comment 15

[djoham]

cc'ing self. Sorry for the spam. This bug just bit me in the butt on a project.
I had to redesign my authentication scheme to get around it... :(

Comment 16

[Darin Fisher]

removing stale nsbeta3 keyword.

Comment 17

[Eric Pollmann]

*** Bug 68666 has been marked as a duplicate of this bug. ***

Comment 18

[Eric Pollmann]

*** Bug 70057 has been marked as a duplicate of this bug. ***

Comment 19

[Eric Pollmann]

The last two dups had between them 5 duplicates.  This seems like a somewhat
serious usability issue in some cases.  Is there any chance this could get fixed
sooner than "Future"?

Comment 20

[Boris Zbarsky [:bzbarsky]]

keyword from bug 70057

Comment 21

[Adam Lock]

Read bug 41241 for a similar problem to do with window targets that I fixed.

It's possible that the form.submit() impl is failing to find or provide the 
correct window target so that the first opened POST channel is disgarded and 
then resubmitted as a normal GET operation.

Comment 22

[Eric Pollmann]

*** Bug 69047 has been marked as a duplicate of this bug. ***

Comment 23

[Darin Fisher]

my changes for bug 76866 will fix this bug as well.

Comment 24

[Darin Fisher]

the http branch didn't completely fix this problem.  we also need to rewind
the post data stream before resending.  attaching a patch that does this.

Comment 25

[Darin Fisher]

Attachment: fixes the problem

Comment 26

[Darin Fisher]

*** Bug 75783 has been marked as a duplicate of this bug. ***

Comment 27

[Bradley Baetz]

r=bbaetz

Comment 28

[Scott MacGregor]

rs=mscott

Comment 29

[Darin Fisher]

fix checked in

Comment 30

[Boris Zbarsky [:bzbarsky]]

*** Bug 81188 has been marked as a duplicate of this bug. ***

Comment 31

[Eric Pollmann]

*** Bug 83519 has been marked as a duplicate of this bug. ***

Comment 32

[Boris Zbarsky [:bzbarsky]]

*** Bug 83094 has been marked as a duplicate of this bug. ***

Comment 33

[Tom Everingham]

verified:
winNT 2001061304
mac os9 2001061308
Linux rh6 2001061308

Comment 34

[benc]

+testcase, thanks for setting that up!

Comment 35

[Boris Zbarsky [:bzbarsky]]

*** Bug 82475 has been marked as a duplicate of this bug. ***