Re: NSS and configure and pkg-config

Hi,
Guenter schrieb:
> will tomorrow give it a try with gnutls - but for NSS it looks already
> great!
configured with:
export PKG_CONFIG_PATH=/usr/i686-pc-mingw32/sys-root/mingw/lib/pkgconfig
./configure --host=i686-pc-mingw32 \
--with-random=notused --with-zlib --enable-sspi --enable-ipv6
--disable-shared --without-ssl --with-gnutls

from config.log:
configure:26300: checking for pkg-config
configure:26319: found /usr/bin/pkg-config
configure:26332: result: /usr/bin/pkg-config
configure:26342: checking for gnutls options with pkg-config
configure:26350: result: found
configure:26404: checking for gnutls_check_version in -lgnutls
configure:26436: i686-pc-mingw32-gcc -o conftest.exe -g0 -O2
-Wno-system-headers  -I/usr/i686-pc-mingw32/sys-root/mingw/include
-lgcrypt -lgpg-error -L/usr/i686-pc-mingw32/sys-root/mingw/lib
conftest.c -lgnutls  -lssh2 -lwldap32 -lws2_32  -lz -lgnutls   >&5
configure:26443: $? = 0
configure:26464: result: yes
configure:26488: detected GnuTLS version 2.8.1
configure:26495: Added /usr/i686-pc-mingw32/sys-root/mingw/lib to
LD_LIBRARY_PATH

and finally the binary:
wine curl.exe -V
curl 7.19.7-CVS (i686-pc-mingw32) libcurl/7.19.7-CVS GnuTLS/2.8.1

Fail to fetch body using curl_easy_perform

Hi,

I faced a problem that I can’t fetch the response body using the
curl_easy_perform.
The same code I used works well before.
But this time, although I can get the Header but the following body
can't be fetched.
(Same payload can be shown correctly using IE/Chrome/Firefox)

As my observation, it seems that if the header and body at the different
packets.
CURL can't parse the body well.
I am not sure if it is the known restriction for the CURL or my
observation is wrong.

I attached a segment of the packets I captured.
Could anyone kindly help me to figure out where the problem is?

Thanks and Best Regards,
YC

========================================================================
0000 00 1f c6 88 54 9c 00 02 d1 09 50 3a 08 00 45 00 ....T.....P:..E.
0010 00 55 87 95 40 00 40 06 1c c0 c0 a8 0a 8f c0 a8 .U.. <at> . <at> .........
0020 0a 6e 00 50 12 82 42 15 4c 24 09 ae 42 7d 50 18 .n.P..B.L$..B}P.
0030 19 20 e4 70 00 00 48 54 54 50 2f 31 2e 31 20 32 . .p..HTTP/1.1 2
0040 30 30 20 4f 4b 0d 0a 43 6f 6e 74 65 6e 74 2d 74 00 OK..Content-t
0050 79 70 65 3a 20 69 6d 61 67 65 2f 6a 70 65 67 0d ype: image/jpeg.
0060 0a 0d 0a

Re: NSS and configure and pkg-config

On Tue, 1 Sep 2009, Guenter wrote:

> All we need now is a libssh2.pc file - seems we currently dont have such, 
> but would be useful since one needs to know if libssh2 is build against 
> either openssl, or libgcrypt ...

I've just now pushed my first take at a libssh2.pc to the libssh2 git repo.

--

-- 

  / daniel.haxx.se

Re: Question regarding curl_multi_perform

On Mon, 31 Aug 2009, subbuscenter-curl <at> yahoo.com wrote:

>>> If the curl_multi_perform() call returns CURLM_OK but
>>> sets running_handles to 0, does it mean that curl is done with all its 
>>> handles and has nothing else to do?
>>
>> Yes.
>
> Thanks. May be the man page for curl_mutli_perform should mention this...

I think it already does. Can you please suggest a better phrasing that makes 
you understand it better than the existing wording?

>>> I also find that if I call curl_multi_info_read(),
>> there is a message, and the status is 0.
>
> I mean the status as set by  the call as follows:
>
> int status;
> cc = curl_easy_getinfo(easy_handle, CURLINFO_RESPONSE_CODE, &status);
> // The value of 'cc' is CURLE_OK, and the value of 'status' is 0.
>
> Can I assume that a 0 status here means that it is an error like ECONNRESET 
> or ECONNREFUSED?

No, that's not really a valid assumption. As the man page explains:

    "This will be zero if no server response code has been received"

It can be due to a variety of reasons. You should use the individual easy 

RE: Include path problem

Guenter wrote:
> Then you probably also have not seen our nice autobuild stats:
> http://curl.haxx.se/auto/

>if you have Perl for your OS400 platform + some spare CPU cycles, just
add it, and we can all see when we break OS400 somehow ...!

Alas, Perl is not supported by IBM on OS400 (there are some unofficial
distributions, but compatibility is weak).
In addition, the script requires some OS400 specific update that I do
not feel able to code: I can read Perl a little, but I don't speak it !

Thanks for the proposal
Patrick

[PATCH] find libssh2 with pkg-config

Hi

As I just made the libssh2 installtion install a pkg-config file, I thought 
I'd also contribute a little fix for curl's configure to find libssh2 using 
that pkg-config approach.

I didn't test this a lot so I'm starting off by mailing it here in the hope 
that someone will find it interesting and try it and tell me how it works.

1) Apply to code
2) run ./buildconf
3) run ./configure --with-libssh2 and see it detect libssh2 properly

Possibly we should make it detect libssh2 even without the explicit option, 
but...

--

-- 

  / daniel.haxx.se

How to determine Server support for range request

Re: How to determine Server support for range request

On Tue, 1 Sep 2009, centrio <at> gmail.com wrote:

> If server does not support range request and we make use of any of above 
> three options, easy_perform retuns saying range_error. I want to determine 
> similar thing, but without making any of these three options. IS it 
> possible? Are there any Http Headers that I can set in the request which 
> will tell me the server behavior??

See http://tools.ietf.org/html/draft-ietf-httpbis-p5-range-07 for all the 
details.

The rough summary is this:

    Origin servers that accept byte-range requests MAY send

      Accept-Ranges: bytes

    but are not required to do so.

--

-- 

  / daniel.haxx.se

Re: Question regarding curl_multi_perform

> >>> If the curl_multi_perform() call returns
> CURLM_OK but
> >>> sets running_handles to 0, does it mean that
> curl is done with all its handles and has nothing else to
> do?
> >> 
> >> Yes.
> > 
> > Thanks. May be the man page for curl_mutli_perform
> should mention this...
> 
> I think it already does. Can you please suggest a better
> phrasing that makes you understand it better than the
> existing wording?

I assume http://curl.haxx.se/libcurl/c/curl_multi_perform.html is the latest man page.
I would prefer to see something explicit, such as "If the call returns 'CURLM_OK', and 'running_handles'
is set to 0, 'libcurl' has completed the transfer for all the easy handles in 'multi_handle'". I may also be
tempted to add "In this case, a call to curl_multi_fdset() will set 'maxfd' to -1 (see curl_multi_fdset())"

> 
> >>> I also find that if I call
> curl_multi_info_read(),
> >> there is a message, and the status is 0.
> > 
> > I mean the status as set by  the call as
> follows:
> > 
> > int status;
> > cc = curl_easy_getinfo(easy_handle,
> CURLINFO_RESPONSE_CODE, &status);
> > // The value of 'cc' is CURLE_OK, and the value of
> 'status' is 0.
> > 
> > Can I assume that a 0 status here means that it is an
> error like ECONNRESET or ECONNREFUSED?
> 
> No, that's not really a valid assumption. As the man page
> explains:
> 
>    "This will be zero if no server response
> code has been received"
> 
> It can be due to a variety of reasons. You should use the
> individual easy transfer's return code to see how its
> transfer did.

What is the call to get the individual easy transfer's return code? I thought it was curl_easy_getinfo
(..CURL_RESPONSE_CODE ..). Or, do you mean CURLINFO_HTTP_CONNECTCODE or CURLINFO_OS_ERRNO?

Thanks for the prompt responses.

-Subbu

Re: tftp - time to break it

On Fri, 14 Aug 2009, koettermarkus <at> gmx.de wrote:

> as the 7.19.6 release is done, we could continue with tftp
>
> Order how I'd do things:

Is anyone working on this or is interested on working on it? We just got 
(another) bug report on the current TFTP functionality not doing fine on 
>32MB files:

http://sourceforge.net/tracker/index.php?func=detail&aid=2848436&group_id=976&atid=100976

--

-- 

  / daniel.haxx.se