Bug 16741 – [GTK] Gmail error

• Component: WebKitGTK
• Status: RESOLVED WORKSFORME

[Tomas Rimkus]

When viewing e-mails on gmail, I get an error, saying: "Oops... the system encountered a problem (#500)".
webkit r29169.
Here is a screenshot:
http://img225.imageshack.us/img225/6369/gmailki1.png

[Luca Bruno]

Yes i can reproduce the problem.

This is what happens:
- Curl sends a POST /mail/?ui.....etc with Content-Length: 19
- The server response is 500, with chunked transfer encoding and text/javascript content type

Seems a cURL bug, but i would bet it's a new Gmail feature. Can i ask you if you ever seen that "Loading..." with yellow background? I never saw it.

[Tomas Rimkus]

(In reply to comment #1) 
> Seems a cURL bug, but i would bet it's a new Gmail feature. Can i ask you if
> you ever seen that "Loading..." with yellow background? I never saw it.
> 

Yellow? No, I don't remember any.

[Tomas Rimkus]

(In reply to comment #2)
> (In reply to comment #1) 
> > Seems a cURL bug, but i would bet it's a new Gmail feature. Can i ask you if
> > you ever seen that "Loading..." with yellow background? I never saw it.
> > 
> 
> Yellow? No, I don't remember any.
> 

Sorry. I sometimes do get those when clicking on Sent, Spam, Drafts, etc.

[Luca Bruno]

Good to hear :)
Why do you think it's a regression then?

[Luca Bruno]

Sorry didn't read your last comment, i mean, do you remember that "Loading" sometimes ago? Not now... to know if it's a new Gmail feature.

[Tomas Rimkus]

(In reply to comment #5)
> Sorry didn't read your last comment, i mean, do you remember that "Loading"
> sometimes ago? Not now... to know if it's a new Gmail feature.
> 

Now that I think, it seems to me that it was there all the time. Anyway, I think the best thing to do would be compiling let's say r28814 to be sure.

[Tomas Rimkus]

Same issue with r28806. That's no regression. Sorry for false alarm.

[Luca Bruno]

No problem, thank you very much :)

I guess it might be related also when getting contacts on the left-side, that always fail.
For both this and contacts, the POST is issued to the web server.
The difference is that for the box on the top, the POST causes the server to return 503, instead the POST for the contacts gets a 404 response.
For debugging purposes, ResourceHandle::loadResourcesSynchronously is not called in Gmail while the two bugs happen.

[Luca Bruno]

Well, maybe i found a possible cause that stop the left-side chat contacts from being loaded.
This is the request with webkit:
GET /mail/channel/test?at=null&VER=5&it=4098&MODE=init&zx=l7gbg4g76i7n&t=1 HTTP/1.1
This one is with mozilla:
GET /mail/channel/test?at=XXXXXXXXXXX&VER=5&it=4098&MODE=init&zx=l7gbg4g76i7n&t=1 HTTP/1.1

[Alp Toker]

Downgrading Priority/Severity to P2/Major. (Right now we preserve P1 for only the most critical crashers and rendering bugs in the GTK+ port.)

[Dirk Schulze]

I got error #611 with the latest soup-trunk (with cookie-fix).

[Duy Nguyen]

libsoup r1148 with WEBKIT_SOUP_LOGGING=1 I got this:

> POST /mail/?ui=2&ik=ff8a75e11f&view=au&rt=j HTTP/1.1
> Soup-Debug-Timestamp: 1217678207
> Soup-Debug: SoupSessionAsync 1 (0x80ebbb8), SoupMessage 72 (0xa537cf8), SoupSocket 5 (0x83b0070)

< HTTP/1.1 411 Length Required
< Soup-Debug-Timestamp: 1217678208
< Soup-Debug: SoupMessage 72 (0xa537cf8)

Libsoup bug?

[Xan Lopez]

*** Bug 25775 has been marked as a duplicate of this bug. ***

[Alexander Butenko]

this should be fixed with libsoup 2.27.2
http://git.gnome.org/cgit/libsoup/commit/?id=74e7c3b0d2201058060233ce2d99860a1c43c646

At least i dont see this error anymore on gmail.

[dx]

I confirm what Alex Butenko said, it works now with that version :D

[Jan Alonzo]

(In reply to comment #15)
> I confirm what Alex Butenko said, it works now with that version :D
> 

It would be nice if we have a regression test for this bug (and bug #26271, though haven't looked if we've already covered this in one of the http/tests/xmlhttprequest/ tests).

[Jan Alonzo]

*** Bug 26271 has been marked as a duplicate of this bug. ***

[dx]

(In reply to comment #16)
> (In reply to comment #15)
> > I confirm what Alex Butenko said, it works now with that version :D
> > 
> 
> It would be nice if we have a regression test for this bug (and bug #26271,
> though haven't looked if we've already covered this in one of the
> http/tests/xmlhttprequest/ tests).

This is not a regression AFAIK (a newer version fixes it, but no older version had this working), and i don't get how would you test it anyway - writing a server that complains when there is content type but no body?

[Jan Alonzo]

Closing as this seemed to be a libsoup bug which is already fixed.