cURL / Mailing Lists / curl-library / Single Mail

curl-library

Re: [PATCH] Better pipelining, round two

From: Dmitry Kurochkin <dmitry.kurochkin_at_gmail.com>
Date: Tue, 15 Jan 2008 17:33:24 +0300

2008/1/15, Daniel Stenberg <daniel_at_haxx.se>:
> On Mon, 14 Jan 2008, Dmitry Kurochkin wrote:
>
> > Here is second version of the patch. I think it fixes all mentioned issues
> > with previous version. Changes:
>
> I'm fine with these changes. One thing that struck me, that I wanted to try
> on you is this:
>
> Shouldn't we better move most (if not all) those pipelining functions from
> lib/url.c to lib/multi.c instead since they're all basically just used from
> there? Then we can make them static and we'll make url.c slightly smaller...
>
> Possibly we could even make a new source file for just the pipelining magic,
> but I'm not sure that'll give us many benefits right now.
>
> Your thoughts on this?

This makes sense to me. I will move functions to multi.c and update
patch when I am back at home.

What else should be done before this can be accepted to CVS?

>
> > ./runtests.pl -t 530 still fails for me. But:
>
> I fails for me too, and it says it fails on 1054 but when I re-run with -t1054
> it runs fine, although I can make it fail repeatedly with -t1053...!

Yes. This seems strange to me that results differ from one run to
another. Also I have different number of mallocs, but I think this
depends on configure options.

Regards,
  Dmitry

>
> But I'm now quite sure this isn't related to your patch at all, since I get
> this problem (at the exact same spot) even without your patch applied! :-/
>
> ("LIMIT multi.c:1418 malloc reached memlimit" with the current CVS code...)
>
> I'll see if I can sort it out...
>
> --
> Commercial curl and libcurl Technical Support: http://haxx.se/curl.html
>
Received on 2008-01-15