Create GTlsConnectionBase, GTlsInputStreamBase, GTlsOutputStreamBase,
which implement the generic "translate from a BSD-sockets-y API to the
GTlsConnection API" semantics, and rebase GTlsConnectionGnutls on top
of that.

It claims to be deprecated, but $GNOME_DESKTOP_SESSION_ID seems to be
the only reliable identifier of GNOME-ness (other than probing D-Bus,
which we can't do from here).

https://bugzilla.gnome.org/show_bug.cgi?id=701377

Detect anything beginning with "gnome". Handles gnome-classic,
and custom gnome-jhbuild or similar sessions.

https://bugzilla.gnome.org/show_bug.cgi?id=700607

automake ends up setting TEST_PROGS to " " in this case, but
Makefile.decl was assuming that if TEST_PROGS is not empty then it
contains actual tests, and then ended up bailing out by passing an no
tests to gtester. Fix it to use $(strip) to get the right result.

Also, drive-by automake warning fix in proxy/tests/Makefile.am

https://bugzilla.gnome.org/show_bug.cgi?id=700628

The pkcs11 tests were using g_assert(!G_IS_OBJECT()) to verify that
objects had been destroyed, but that's just crazy, and will crash
sometimes. Fix them to do a ref_count check instead.

Iain Lane authored May 14, 2013 and Dan Winship committed May 14, 2013

https://bugzilla.gnome.org/show_bug.cgi?id=700286

Fix the handling of SOCKS proxies with g_proxy_resolver_lookup_async()
and the libproxy resolver. (Also, make the sync proxy resolver
properly cancellable.)

Add some tests of the libproxy resolver, based on the existing gnome
proxy resolver tests. Unfortunately, libproxy doesn't implement ignore
hosts in exactly the same way GProxyResolverGnome does, so we need to
keep track of which URIs they give different results for...

https://bugzilla.gnome.org/show_bug.cgi?id=699359

Iain Lane authored Apr 29, 2013 and Dan Winship committed Apr 30, 2013

This avoids falling back to libproxy unnecessarily.

https://bugs.launchpad.net/ubuntu/+source/glib-networking/+bug/1172792

https://bugzilla.gnome.org/show_bug.cgi?id=698936

Automatic proxy resolution mode was accidentally getting more-or-less
ignored and treated as "none".

Also add a test to prevent regressions, this test tries to do a async handshake,
and waits for the actual handshake to finish, but before allowing the callback to happen
it tries to do a write.

https://bugzilla.gnome.org/show_bug.cgi?id=697754

If a server requests a rehandshake but then receives application data,
gnutls 2.x silently buffers that data, but gnutls 3.x returns a
(non-fatal) error. Do the buffering inside GTlsConnectionGnutls so
that 2.x and 3.x end up behaving the same.

https://bugzilla.gnome.org/show_bug.cgi?id=695062

Emmanuel Pacaud authored Mar 25, 2013 and Dan Winship committed Apr 03, 2013

(recommit with correct bug url)

https://bugzilla.gnome.org/show_bug.cgi?id=696519

This reverts commit 98021800.

Emmanuel Pacaud authored Mar 25, 2013 and Dan Winship committed Apr 03, 2013

https://bugzilla.gnome.org/show_bug.cgi?id=692134

The cancellable pollfd was overwriting the waiting_op one

https://bugzilla.gnome.org/show_bug.cgi?id=696881