[Webkit-unassigned] [Bug 49018] [GTK] response.isNull() assert when using directory file URI

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=49018





--- Comment #23 from Sergio Villar Senin <svillar at igalia.com>  2010-11-11 03:38:02 PST ---
(In reply to comment #22)
> Created an attachment (id=73535)
 --> (https://bugs.webkit.org/attachment.cgi?id=73535&action=review) [details]
> Fix assert using directory file URI
> 
> Updated patch.

The patch looks good to me, don't know if we need such amount of new asserts tough :). Being a bit picky I wouldn't say "Ensure a response for protocol using GIO.", we're ensuring it for any protocol not directly handled by WebKitGtk+, it could be gio, libsoup or whatever.

And regarding the comment in the ChangeLog I suggest something more descriptive like:
"Do ensure that didReceiveResponse happens before any call to didReceiveData. That was not true for file: URIs pointing to directories and thus was triggering an assertion." I think that's a bit better because your patch is really fixing that instead of only the directory issue mentioned in the bug. Just my 2 cents.

Thx!

> 
> Code:
> - Added more assert()
> - Moved readCallback() assert lower since sending a response is not required if read failed.
> - Use d->m_response for data:// to avoid asserting on !d->m_response.isNull() later
> - Using handle->shouldContentSniff() along with d->m_response.isNull() to decide if a response should be sent in sendRequestCallback()
> 
> Test:
> - renamed test to crash-display-local-directory.html
> - Creating iframe in HTML and not using JS
> - Using iframe onload() instead of body onload()
> - Using <p> instead of <div> and <br/>

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.