Daniel Stenberg wrote:
> On Thu, 6 May 2010, Howard Chu wrote:
>
> I really like the spirit of this patch. It makes a lot of sense to me as it
> moves the send/recv functions out to the code that deal with the particular
> protocols.
>
> I'm posting my updated version of your work here mostly because I'm heading
> off to bed now and it's not quite working yet.
>
> The one current flaw I've found I believe is this:
>
> Curl_read() calls conn->recv() and if that returns -1, Curl_read() returns
> the curlcode.
>
> This is bad since Curl_read() used to return -1 for EAGAIN cases and now it
> returns CURLE_AGAIN instead so for example lib/pingpong.c:Curl_pp_readresp()
> doesn't treat the return code from Curl_read() correctly.
>
> It should be fairly easy to fix and I'll give it another shot tomorrow unless
> someone else gets there first.

I've chased down all uses of Curl_read and Curl_read_plain and updated them.
This diff is based on your posted patch.

-- 
   -- Howard Chu
   CTO, Symas Corp.           http://www.symas.com
   Director, Highland Sun     http://highlandsun.com/hyc/
   Chief Architect, OpenLDAP  http://www.openldap.org/project/

-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette: http://curl.haxx.se/mail/etiquette.html