Bugfix: https handshake fails using gnutls 3 on windows
http://sourceforge.net/tracker/index.php?func=detail&aid=3441084&group_id=976&atid=100976
New gnutls versions have an error handler that knows about Winsock
errors, which is why gnutls_transport_set_global_errno() was deprecated
and then removed.
This is a correction of commit f5bb370 (blame me) which meant to
reimplement gnutls_transport_set_global_errno(), which is not necessary.
Use gnutls_priority_set_direct() instead of gnutls_protocol_set_priority().
Remove the gnutls_certificate_type_set_priority() use since x509 is the
default certificate type anyway.
Reported by: Vincent Torri
The default lowat level for gnutls-2.12* is set to zero to avoid
unnecessary system calls and the gnutls_transport_set_lowat function has
been totally removed in >=gnutls-3 which causes build failures.
Therefore, the function shouldn't be used except for versions that
require it, <gnutls-2.12.0.
On second thought, I think CURLE_TLSAUTH_FAILED should be eliminated. It
was only being raised when an internal error occurred while allocating
or setting the GnuTLS SRP client credentials struct. For TLS
authentication failures, the general CURLE_SSL_CONNECT_ERROR seems
appropriate; its error string already includes "passwords" as a possible
cause. Having a separate TLS auth error code might also cause people to
think that a TLS auth failure means the wrong username or password was
entered, when it could also be a sign of a man-in-the-middle attack.
Instead of polluting many places with #ifdefs, we create a single place
for this function, and also check return code properly so that a NULL
pointer returned won't cause problems.
On Windows, translate WSAGetLastError() to errno values as GNU
TLS does it internally, too. This is necessary because send() and
recv() on Windows don't set errno when they fail but GNU TLS
expects a proper errno value.
Bug: http://curl.haxx.se/bug/view.cgi?id=3110991
When no timeout is set, we call the socket_ready function with a timeout
value of 0 during handshake, which makes it loop too much/fast in this
function. It also made this function return CURLE_OPERATION_TIMEDOUT
wrongly on a slow handshake.
However, the particular bug report that highlighted this problem is not
solved by this fix, as this fix only makes the more proper error get
reported instead.
Bug: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=594150
Reported by: Johannes Ernst
FTP(S) use two connections that can be set to different recv and
send functions independently, so by introducing recv+send pairs
in the same manner we already have sockets/connections we can
work with FTPS fine.
This commit fixes the FTPS regression introduced in change d64bd82.
Howard Chu brought the bulk work of this patch that properly
moves out the sending and recving of data to the parts of the
code that are properly responsible for the various ways of doing
so.
Daniel Stenberg assisted with polishing a few bits and fixed some
minor flaws in the original patch.
Another upside of this patch is that we now abuse CURLcodes less
with the "magic" -1 return codes and instead use CURLE_AGAIN more
consistently.
The recent overhaul of the SSL recv function made this treat a
zero returned from gnutls_record_recv() as an error, and this
caused our HTTPS test cases to fail. We leave it to upper layer
code to detect if an EOF is a problem or not.
should introduce an option to disable SNI, but as we're in feature freeze
now I've addressed the obvious bug here (pointed out by Peter Sylvester): we
shouldn't try to enable SNI when SSLv2 or SSLv3 is explicitly selected.
Code for OpenSSL and GnuTLS was fixed. NSS doesn't seem to have a particular
option for SNI, or are we simply not using it?
in the gnutls code where we were checking for negative values for errors,
when the man pages state that GNUTLS_E_SUCCESS is returned on success and
other values indicate error conditions.
curl_global_init() function to properly maintain the performing functions
thread-safe. We've previously (28 April 2007) moved the init to a later time
just to avoid it to fail very early when libgcrypt dislikes the situation,
but that move was bad and the fix should rather be in libgcrypt or
elsewhere.
(http://curl.haxx.se/bug/view.cgi?id=2255627) which pointed out that a
program using libcurl's multi interface to download a HTTPS page with a
libcurl built powered by OpenSSL, would easily get silly and instead hand
over SSL details as data instead of the actual HTTP headers and body. This
happened because libcurl would consider the connection handshake done too
early. This problem was introduced at September 22nd 2008 with my fix of the
bug #2107377
The correct fix is now instead done within the GnuTLS-handling code, as both
the OpenSSL and the NSS code already deal with this situation in similar
fashion. I added test case 560 in an attempt to verify this fix, but
unfortunately it didn't trigger it even before this fix!
(http://curl.haxx.se/bug/view.cgi?id=2154627) which pointed out that libcurl
uses strcasecmp() in multiple places where it causes failures when the
Turkish locale is used. This is because 'i' and 'I' isn't the same letter so
strcasecmp() on those letters are different in Turkish than in English (or
just about all other languages). I thus introduced a totally new internal
function in libcurl (called Curl_ascii_equal) for doing case insentive
comparisons for english-(ascii?) style strings that thus will make "file"
and "FILE" match even if the Turkish locale is selected.
better control at the exact state of the connection's SSL status so that we
know exactly when it has completed the SSL negotiation or not so that there
won't be accidental re-uses of connections that are wrongly believed to be
in SSL-completed-negotiate state.
such as the CURLOPT_SSL_CTX_FUNCTION one treat that as if it was a Location:
following. The patch that introduced this feature was done for 7.11.0, but
this code and functionality has been broken since about 7.15.4 (March 2006)
with the introduction of non-blocking OpenSSL "connects".
It was a hack to begin with and since it doesn't work and hasn't worked
correctly for a long time and nobody has even noticed, I consider it a very
suitable subject for plain removal. And so it was done.
Yang Tse
Mark Brand
Martin Storsjo
Daniel Stenberg
Tim Harder
Quinn Slack
Matthias Bolte
Howard Chu
Jerome Vouillon
Kamil Dudka
Gunter Knauf
Michal Marek
Dan Fandrich