82050 - XMLHttpRequest send() must ignore argument if open() method is GET

Status: VERIFIED FIXED

(Core :: XML, defect)

Description

[Heikki Toivonen (remove -bugzilla when emailing directly)]

Any request to load data from a server is answered with either

403 Forbidden

or

405 Method Not Allowed

The former error is returned when the script tried to POST to a script in
cgi-bin directory, and the latter when the script tried to GET an XML file in
the normal HTML document area on the webserver.

I have not made any changes to the XML Extras component in this time. Others
have modied XMLExtras in this time, but I do not see anything in there that
could have caused this (it is still possible). There have been lots of changes
in necko, though, and xpcdom has landed (though at least initially after the
landing XMLHttpRequest did work).

Things are even worse in embedding land. The testcase in URL at least pops up
alert boxes in Browser, but winEmbed and mfcEmbed do not do that. Alerts work in
embedding, though. This could be a separate bug.

I checked with pollmann what the webserver is actually seeing, and it looks like
we might be giving the server a wrong request. Here is what the server saw last
Wednesday when I tried to do "GET http://blueviper/heikki/tupshin4.xml":

Client1> POST /tupshin4.xml HTTP/1.1
Client1> Host: heikki
Client1> User-Agent: Mozilla/5.0 (Windows; U; WinNT4.0; en-US; rv:0.9+) Gecko/20
010514
Client1> Accept: */*
Client1> Accept-Language: en
Client1> Accept-Encoding: gzip,deflate,compress,identity
Client1> Accept-Charset: ISO-8859-1, utf-8; q=0.667, *; q=0.667
Client1> Keep-Alive: 300
Client1> Connection: keep-alive
Client1> Content-type: text/xml
Client1> Content-Length: 13
Client1> Client1> <test></test>ServerClient1> HTTP/1.1 404 Not Found

Comment 1

[Heikki Toivonen (remove -bugzilla when emailing directly)]

I can not be 100% certain about the validity of experiment I did with pollmann,
but something fishy is going on somewhere. Maybe necko changed subtly but
XMLHttpRequest was not changed along with that. Or maybe some recent xpcdom fix
did XMLHttpRequest in.

My first guess is necko changes, Cc:ng darin, gagan and dougt.

I have marked this blocking bug 73958 which I am 75% sure is some kind of off by
one error in some stream implementation (cache stream bad?) or some other
weirdness. This bug, bug 73958 and maybe bug 82032 are the ones we would need
for 0.9.1.

You could argue that I should mark all XMLHttpRequest bugs depending on this,
but I feel lazy...

Tomorrow I'll start downloading nightly builds to see when things broke exactly.

Comment 2

[Darin Fisher]

the http re-arch branch landed friday, 5/11... and, this may be a regression
due to that.  you might try using a packet tracing tool like ngrep to capture
activity on port 80 with today's build as well as an older build that works.
if you could attach these logs to this bug report i might be able to help
diagnose the problem (if it happens to be with necko).

Comment 3

[Taras Tielkes]

(CCing)
A nice Windows tool is: 
http://www.pocketsoap.com/tcptrace/default.asp

Comment 4

[Heikki Toivonen (remove -bugzilla when emailing directly)]

Hmm.. it seems part of this was due to the test server configuration changing.
Using the browser the normal post.html and syncpost.html tests work.

GET does not work (the testcase in bug 73958).

Using winEmbed async does not work, but sync does. The world turned upside down.
Sync should not work in embedding 'cos there is no implementation for it! Or so
I thought until recently...

Comment 5

[Heikki Toivonen (remove -bugzilla when emailing directly)]

The GET method, and what is blocking testing of bug 73958 broke between builds
051104 and 051208 (the former works, the latter doesn't).

This bonsai query shows the checkins between 051102 (assuming 2 hours for pull
and build) and 051208:

http://bonsai.mozilla.org/cvsquery.cgi?treeid=default&module=SeaMonkeyAll&branch=HEAD&branchtype=match&dir=&file=&filetype=match&who=&whotype=match&sortby=Change+Size&hours=2&date=explicit&mindate=05%2F11%2F2001+02%3A00%3A00&maxdate=05%2F12%2F2001+08%3A00%3A00&cvsroot=%2Fcvsroot

There are lots of checkins, but the most suspicuous one is HTTP branch landing
in my opinion. Changes in xmlextras directory are limited to HTTP header
handling (which kinda points in the direction of this bug also).

I have also noticed something weird going on with the testcase for bug 73958.
When I open tupshin.xul document mork (history database?) seems to get corruped.
Maybe that happens because of the wrong headers, no sure...

I still haven't found out what is going on with the embedded builds, the nightly
builds I have tested all had winEmbed working strange. I should probably open a
new bug on that...

Comment 6

[Heikki Toivonen (remove -bugzilla when emailing directly)]

Ok, so we actually are not touching the changed header code in xmlextras (placed
breakpoints there).

But I did notice this line being commented out:

!   //mChannel->SetAuthTriedWithPrehost(authp);

and what we are getting in 405 Method Not Allowed from the server (POST method
even though we asked GET). Maybe that line has nothing to do with it...

Let's see if we can come up with ngrep output...

Comment 7

[Darin Fisher]

I talked to vidur about that line back when I was making the http changes.
Turns out it isn't needed... it basically just says that an authentication
dialog should always be displayed even if the URL contains a user:pass already.

I don't see how this could have anything to do with the problem described here.

The request method will become POST if you set an upload stream on
nsIHttpChannel.  Is this happening?

Comment 8

[Heikki Toivonen (remove -bugzilla when emailing directly)]

Allright, I have at least a workaround. Now it kinda even makes sense why the
tescase failed, but I don't know why it worked before the HTTP landing.

Maybe necko did not change the method? Should I close this bug as INVALID (wrong
usage pattern)? If wrong usage, I guess I might need to throw an error in the
send method to inform the developer that what they are trying to do does not
make sense...

The testcase was doing

[XMLHttpRequest object].open("GET","http://green/heikki/tupshin4.xml");
[XMLHttpRequest object].send([dummy XML DOM document]);

So this kinda looks like we are trying to do a GET but sending a serialized
dummy XML document to the server at the same time.

The workaround is changing the second line to:

[XMLHttpRequest object].send(null);


Before the HTML landing things still somehow worked. This is ngrep output:

T 208.12.36.155:4354 -> 208.12.36.110:80 [AP]
  GET /heikki/tupshin4.xml HTTP/1.1..Host: green..User-Agent: Mozilla/5.0 (Wi
  ndows; U; WinNT4.0; en-US; rv:0.9+) Gecko/20010511..Accept: */*..Accept-Lan
  guage: en..Accept-Encoding: gzip,deflate,compress,identity..Accept-Charset:
   ISO-8859-1, utf-8; q=0.667, *; q=0.667..Keep-Alive: 300..Connection: keep-
  alive....
##
U 208.12.36.67:1052 -> 208.12.36.255:135
  ........3;..............3;.i.............%]vM\.....$C......................
  .....
#
T 208.12.36.110:80 -> 208.12.36.155:4354 [A]
  HTTP/1.1 200 OK..Date: Thu, 24 May 2001 06:02:51 GMT..Server: Apache/1.3.19
   (Unix)  (Red-Hat/Linux) mod_ssl/2.8.1 OpenSSL/0.9.6 DAV/1.0.2 PHP/4.0.4pl1
   mod_perl/1.24_01..Last-Modified: Thu, 17 May 2001 00:10:45 GMT..ETag: "320
  2d8-a87-3b031705"..Accept-Ranges: bytes..Content-Length: 2695..Keep-Alive:
  timeout=15, max=100..Connection: Keep-Alive..Content-Type: text/xml....<roo
  t>.    <child>This is a child node</child>.    <child>This is a child node<
  /child>.    <child>This is a child node</child>.    <child>This is a child
  node</child>.    <child>This is a child node</child>.    <child>This is a c
  hild node</child>.    <child>This is a child node</child>.    <child>This i
  s a child node</child>.    <child>This is a child node</child>.    <child>T
  his is a child node</child>.    <child>This is a child node</child>.    <ch
  ild>This is a child node</child>.    <child>This is a child node</child>.
    <child>This is a child node</child>.    <child>This is a child node</chil
  d>.    <child>This is a child node</child>.    <child>This is a child node<
  /child>.    <child>This is a child node</child>.    <child>This is a child
  node</child>.    <child>This is a child node</child>.    <child>This is a c
  hild node</child>.    <child>This is a child node</child>.    <child>This i
  s a child node</child>.    <child>This is a child node</child>.    <child>T
  his is a child node</child>.    <child>This is a child node</child>.    <ch
  ild>This is a child node</child>.
#
T 208.12.36.110:80 -> 208.12.36.155:4354 [A]
    <child>This is a child node</child>.    <child>This is a child node</chil
  d>.    <child>This is a child node</child>.    <child>This is a child node<
  /child>.    <child>This is a child node</child>.    <child>This is a child
  node</child>.    <child>This is a child node</child>.    <child>This is a c
  hild node</child>.    <child>This is a child node</child>.    <child>This i
  s a child node</child>.    <child>This is a child node</child>.    <child>T
  his is a child node</child>.    <child>This is a child node</child>.    <ch
  ild>This is a child node</child>.    <child>This is a child node</child>.
    <child>This is a child node</child>.    <child>This is a child node</chil
  d>.    <child>This is a child node</child>.    <child>This is a child node<
  /child>.    <child>This is a child node</child>.    <child>This is a child
  node</child>.    <child>This is a child node</child>.    <child>This is a c
  hild node</child>.    <child>This is a child node</child>.    <child>This i
  s a child node</child>.    <child>This is a child node</child>.    <child>T
  his is a child node</child>.    <child>This is a child node</child>.    <ch
  ild>This is a child node</child>.    <child>This is a child node</child>.
    <child>This is a child node</child>.    <child>This is a child node</chil
  d>.    <child>This is a child node</child>.    <child>This is a child node<
  /child>.    <child>This is a child node</child>.    <child>This is a child
  node</child>.    <child>This is a c
##
T 208.12.36.110:80 -> 208.12.36.155:4354 [AP]
  hild node</child>.    <child>This is a child node</child>.    <child>This i
  s a child node</child>.    <child>This is a child node</child>.</root>.


With current code ngrep shows:

T 208.12.36.155:2743 -> 208.12.36.110:80 [AP]
  POST /heikki/tupshin4.xml HTTP/1.1..Host: green..User-Agent: Mozilla/5.0 (W
  indows; U; WinNT4.0; en-US; rv:0.9+) Gecko/20010514..Accept: text/xml;q=1,
  text/html;q=0.9, image/png;q=1, image/jpeg;q=1, image/gif;q=0.9, text/plain
  ;q=0.8, text/css;q=1, */*;q=0.01..Accept-Language: en..Accept-Encoding: gzi
  p,deflate,compress,identity..Accept-Charset: ISO-8859-1, utf-8; q=0.667, *;
   q=0.667..Keep-Alive: 300..Connection: keep-alive..
##
T 208.12.36.155:2743 -> 208.12.36.110:80 [AP]
  Content-type: text/xml..Content-Length: 13....<test></test>..
##
U 208.12.36.110:800 -> 205.217.237.150:2049
  ..|6...........................(........green.mcom.com.....0...0.......0...
  ........ EZ0..... ....0ZE..........[...H.....public..
#
U 205.217.237.150:2049 -> 208.12.36.110:800
  ..|6........................... EZ0..... ....c7;*[........[...H............
  .......)..........................................c7;;.M.....:...%..h:...%.
  .h..................)...........0.......0......................0ZE;.;.#.=.;
  .:.6.v.;.:.6.v.
#
U 208.12.36.110:800 -> 205.217.237.150:2049
  ..|6...........................(........green.mcom.com.....0...0.......0...
  ........ EZ0..... ....c7;*[........[...H.....work
#
U 205.217.237.150:2049 -> 208.12.36.110:800
  ..|6........................... EZ0..... ....0....&.......[...H............
  .......)..........................................0..;.M.3...;.../.uh;.../.
  uh..................)..........................................c7;;.M.....:
  ...%..h:...%..h
#
U 208.12.36.110:800 -> 205.217.237.150:2049
  ..|6...........................(........green.mcom.com.....0...0.......0...
  ........ EZ0..... ....0....&.......[...H.....heikki..
#
U 205.217.237.150:2049 -> 208.12.36.110:800
  ..|6........................... EZ0..... ....n............[...H............
  .......)..........................................n..;...1..(;...0.8.;..$.u
  *...................)..........................................0..;.M.3...;
  .../.uh;.../.uh
#
U 208.12.36.110:800 -> 205.217.237.150:2049
  ..|6...........................(........green.mcom.com.....0...0.......0...
  ........ EZ0..... ....n............[...H.....tupshin4.xml
#
U 205.217.237.150:2049 -> 208.12.36.110:800
  ..|6........................... EZ0..... ....2...C.1......[...H............
  .......)..........................................2..;...0...;.......;..=;.
  7h..................)..........................................n..;...1..(;
  ...0.8.;..$.u*.
#
T 208.12.36.110:80 -> 208.12.36.155:2743 [AP]
  HTTP/1.1 405 Method Not Allowed..Date: Thu, 24 May 2001 05:42:09 GMT..Serve
  r: Apache/1.3.19 (Unix)  (Red-Hat/Linux) mod_ssl/2.8.1 OpenSSL/0.9.6 DAV/1.
  0.2 PHP/4.0.4pl1 mod_perl/1.24_01..Allow: GET, HEAD, OPTIONS, TRACE..Keep-A
  live: timeout=15, max=100..Connection: Keep-Alive..Transfer-Encoding: chunk
  ed..Content-Type: text/html; charset=iso-8859-1....134..<!DOCTYPE HTML PUBL
  IC "-//IETF//DTD HTML 2.0//EN">.<HTML><HEAD>.<TITLE>405 Method Not Allowed<
  /TITLE>.</HEAD><BODY>.<H1>Method Not Allowed</H1>.The requested method POST
   is not allowed for the URL /heikki/tupshin4.xml.<P>.<HR>.<ADDRESS>Apache/1
  .3.19 Server at green.mcom.com Port 80</ADDRESS>.</BODY></HTML>...0....

Comment 9

[Darin Fisher]

yeah... i believe the old code didn't automatically change the method.  it makes
sense though that if you are going to set an upload stream that it would have
to be a POST (or PUT) method, otherwise we'd simply ignore the upload stream.

Comment 10

[Heikki Toivonen (remove -bugzilla when emailing directly)]

I am marking this INVALID, because the usage that prompts this is so weird (open
with GET method but try to send data to server while getting a file). I may end
up throwing an error in the send() method if I detect this.

Comment 11

[Taras Tielkes]

Hi Heikki,

Couldn't you just ignore the pagemeter altogether if the method is "GET"?

IIRC, that's what ie5+ does.

Comment 12

[Heikki Toivonen (remove -bugzilla when emailing directly)]

Ok, I'll go with that.

Comment 13

[Heikki Toivonen (remove -bugzilla when emailing directly)]

I'll just do a little cleanup... Sorry for spam.

Comment 14

[Heikki Toivonen (remove -bugzilla when emailing directly)]

The fix for this went in with bug 73958 on trunk and 0.9.1 branch.

Comment 15

[Chris Petersen]

Marking verified in the June 04 trunk and branch builds.