[Webkit-unassigned] [Bug 48002] network-simulator.php makes for very slow layout tests

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=48002





--- Comment #4 from Michael Nordman <michaeln at google.com>  2010-10-20 14:07:17 PST ---
(In reply to comment #3)
> > Yes, I see what it's doing and why. The fact remains that we have tests that timeout due to this.
> 
> Can that be fixed in chromium networking code? I don't think CFNetwork has any performance issues with redirects.

It functions fine, its just slow on this non-sensical content. I really don't want to modify the networking code for this.

> 
> > I'm looking thru the appcache tests and noticing that simulated network errors are used thruout, but haven't really noticed (m)any tests that simulate a server error.
> 
> Yeah, we need more of the latter kind.

I'd like to switch some of the former to be some of the latter, and retain some 'network error' cases in there too, but lean on them less.

> > Also, I question whether this is as good of a simulation of a network error as can be constructed. I wonder if we can close the underlying connection from the server-side?
> 
> I couldn't find a better way when making those tests. If there is a way to force Apache to rudely drop a connection, that would be a much better approximation indeed.

I don't know php very well, I didn't see an obvious way to close the connection? If I could get a reference to the underlying 'socket' there is a call to close the socket. 

Another approach could be to simulate a 'network error' could be with a malformed http message, maybe an excessively large header? I'll see if php lets me form a malformed header? It'd be nice if php let me produce the output w/o doing anything 'for me' in this case.

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