[webkit-gtk] webkit2gtk 2.7.4 sending cookie from cache after a 304 with missing Content-Type ?

[Jérémy Lal]

2015-03-25 11:38 GMT+01:00 Jérémy Lal <kapouer at melix.org>:

> 2015-03-25 11:29 GMT+01:00 Carlos Garcia Campos <cgarcia at igalia.com>:
> > El mié, 25-03-2015 a las 11:20 +0100, Jérémy Lal escribió:
> >> 2015-03-25 9:01 GMT+01:00 Carlos Garcia Campos <cgarcia at igalia.com>:
> >> > El mar, 24-03-2015 a las 22:48 +0100, Jérémy Lal escribió:
> >> >> 2015-03-24 21:21 GMT+01:00 Jérémy Lal <kapouer at melix.org>:
> >> >> > I'm seeing something very weird in a complicated setup here, and
> am trying to
> >> >> > understand what's happening:
> >> >> >
> >> >> > 1 - open page urlA, Set-Cookie CA
> >> >> > 2 - do xhr requests to urlB (ETag set in response)
> >> >> > 3 - load same page urlA, Set-Cookie CB
> >> >> > 4 - do xhr request to urlB 304 (If-None-Match hit), a response is
> sent
> >> >> > from expressjs server, without a content-type header set (the
> server
> >> >> > gets Cookie CB)
> >> >>
> >> >> I think i found how to reproduce with this precision in step 4:
> >> >>
> >> >> 4 - do xhr request to urlB 304 (If-None-Match hit)
> >> >> ** with the same Last-Modified header as in step 2 **
> >> >>
> >> >> It makes webkit2gtk 2.7.4 go back in time and it start sending
> cookies
> >> >> from the past !
> >> >>
> >> >>
> >> >> > 5 - a second xhr request to never requested urlC is done: the
> server
> >> >> > gets Cookie CA !
> >> >>
> >> >> > am going to test against webkit2gtk 2.8.0
> >> >>
> >> >> building...
> >> >
> >> > What libsoup version are you using?
> >>
> >> The one from debian jessie, 2.48.
> >> I just gave a try with webkit2gtk 2.8.0 and it does the same.
> >
> > I asked because the new libsoup version 2.50 has several fixes in the
> > disk cache.
>
> I just read the changelog and am almost already convinced the bug is
> already fixes there :)
>
> >
> >> The symptoms are a bit different. Sometimes when i do a test, the bug
> >> does not appear,
> >> and two third of the times it appears.
> >
> > Could you file a bug report with instructions to reproduce the issue? or
> > even better a test case if possible?
>
> I'll try to trim the code down to the minimum. Expect nodejs code...
>


Hi,

i opened
https://bugs.webkit.org/show_bug.cgi?id=145734

with code for reproducing the problem i'm seeing - which is simpler to
understand now :)

Jérémy.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.webkit.org/pipermail/webkit-gtk/attachments/20150607/c83b7e76/attachment.html>