Hi Daniel...

Here is a second installment for a connect_only patch.

Thanks,

...Karl

>From: Daniel Stenberg
>Subject: Re: Passing raw data through a proxy
>Date: Wed, 8 Feb 2006 00:36:30 +0100 (CET)
>
>On Mon, 6 Feb 2006, Karl M wrote:
>
>>Here is a first installment for a connect_only patch. Please take a look
>>at it and comment.
>
>...
>
>>Perhaps CURLINFO_LASTSOCKET would be better that CURLINFO_SOCKET, to
>>emphasize that it is the most recently used socket for the session handle.
>
>Yes, I think it would make sense to have that somewhat clearer by stating
>it in the name itself. I would also appreciate some docs for the options,
>either as nroff patches or as plain texts that I can insert into the nroff
>files manually.
>
>On the patch itself, I have two nits:
>
>o It doesn't apply. It looks like it was line wrapped at a couple of places
> and some lines starts at column 0 which makes them illegal diff
>format...
>
>o Don't use INVALID_SOCKET to indicate a bad socket (since it isn't
> available on all systems), use CURL_SOCKET_BAD. Or... Hm, that's what
>is
> returned to the user if there's no socket available. Now that is a fun
> little excersize. Perhaps just returning -1 there (since it after all is
>a
> 'long' and not a socket type) would make most sense.
>
>In general it looks fine to me, but I'd really like to apply it first and
>read some of the code with the code applied before I comment any further.
>
>--
> Commercial curl and libcurl Technical Support: http://haxx.se/curl.html