| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Even if the caller does not specify a GCancellable the GTask must use the
one that comes with the item by default.
This also sets check_cancellable() to FALSE in the session's GTask in order
not to override error messages previously set on the GTask in case of
cancellations.
https://bugzilla.gnome.org/show_bug.cgi?id=692310
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
This broke CalDAV in evolution. I'm not sure exactly why, but if it
broke that, it might break other things too, so let's just revert
this.
This reverts commit 38901ca5e684a8fac75e6ff740d45c82dd44181a.
https://bugzilla.gnome.org/show_bug.cgi?id=692149
|
| |
|
|
|
|
| |
Signed-off-by: Gheyret Kenji <gheyret@gmail.com>
|
|
|
|
|
|
|
|
|
| |
We must update the priv->eof field in the skip() operation as we do in
read(). This bug was causing redirects not to finish because the
SoupBodyInputStream was creating a pollable source that will never issue
anything instead of a plain timeout source.
https://bugzilla.gnome.org/show_bug.cgi?id=692026
|
|
|
|
|
|
|
|
| |
If g_tls_file_database_new() failed for any reason but
G_TLS_ERROR_UNAVAILABLE, we don't create a fallback DB and so tlsdb stays
NULL.
https://bugzilla.gnome.org/show_bug.cgi?id=691930
|
| |
|
| |
|
| |
|
|
|
|
|
|
| |
SoupRequestHTTP is now implemented in terms of SoupSession SoupMessage
operations, so we might as well just expose those directly, for people
who want streaming APIs, but not SoupRequest.
|
|
|
|
|
|
| |
Remove some includes that I had commented out to see if they were
still needed, but then forgot to actually remove when it turned out
they weren't.
|
|
|
|
|
| |
All supported HTTP auth methods use both username and password, so
assume a password of "" if none is given
|
|
|
|
|
|
|
| |
Calling soup_auth_authenticate() on an already-authenticated
SoupAuthDigest would leak the old username and hashed password.
https://bugzilla.gnome.org/show_bug.cgi?id=690142
|
| |
|
|
|
|
|
|
|
|
| |
We need to signal conn_cond when removing an item from the queue,
since flush_queue() expects that... (and also, since in some cases,
removing a message from the queue will allow another one to start).
https://bugzilla.gnome.org/show_bug.cgi?id=691399
|
| |
|
|
|
|
| |
The po/Makefile.in.inis generated by intltoolize
|
| |
|
|
|
|
|
|
|
|
|
| |
SoupMessage:tls-certificate and SoupMessage:tls-errors were only
getting set for successful https connections. It is useful to have
them be set on failed ones as well. Fix that, and make ssl-test test
it.
https://bugzilla.gnome.org/show_bug.cgi?id=690176
|
|
|
|
|
|
|
|
|
|
| |
Disconnecting clears priv->sock, and in some cases, the caller may
want to look at that after the failed connection. So switch this back
to having the session do the disconnecting itself (which is simpler
now due to various SoupSession rewrites), and remove one inconsistent
case where the connection was disconnecting itself before.
This reverts commit 21d707d66460233596a04f6dc99c945d2dbc3085.
|
| |
|
|
|
|
|
|
|
|
| |
Split the "this message is likely to be requeued" logic into a single
function that the sync and async codepaths can share, and fix it to
not think that SOUP_MESSAGE_NO_REDIRECT messages will be redirected,
or that messages that disable SOUP_TYPE_AUTH_MANAGER will be
authenticated.
|
|
|
|
|
|
|
|
|
| |
GNOME_MAINTAINER_MODE_DEFINES apparently sets
LIBSOUP_DISABLE_DEPRECATED, which is now causing various jhbuild
modules to break. "Fix" this by renaming the define (which is only
still there because gtk-doc complains if it's not). The *right* way to
get deprecation warnings/errors now is to define
SOUP_VERSION_MIN_REQUIRED.
|
|
|
|
|
|
|
| |
Very belatedly, make SoupAuthManager thread-safe.
It also ensures that any SoupAuth that it manages is only manipulated
from one thread at a time.
|
|
|
|
|
|
| |
Add soup_auth_manager_use_auth(), for "preloading" authentication, and
make the old automatically-request-NTLM behavior happen only for
the legacy SoupSession subclasses.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Add a new SoupConnectionAuth class to help with connection tracking,
and make SoupAuthNTLM a subclass of it. Allow a single SoupAuthNTLM to
carry state information about multiple connections.
Make SoupSession store the SoupConnection a SoupMessage is associated
with on the message, and use that from SoupConnectionAuth rather than
tracking sockets by hand like SoupAuthManager had previously done.
Remove the connection tracking in SoupAuthManager, since it is no
longer needed.
|
|
|
|
|
|
| |
Use a single boolean to track whether we're using single-sign-on
(/usr/bin/ntlm_auth) rather than having two parallel sets of
SoupNTLMStates.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
We need to set SOUP_NTLM_AUTH_DEBUG to an empty string (to tell
SoupAuthNTLM to always use built-in NTLM), not leave it unset, which
will let SoupAuthNTLM use the real ntlm_auth binary.
(The effect of this bug was that instead of always using built-in
NTLM, ntlm-test-helper would first try to use ntlm_auth, and fail
because there were no cached credentials, and then fall back to
built-in NTLM, which meant that we were testing the fallback codepath
even in the supposed-to-be-non-fallback case. Unless your username was
"alice" or "bob" and you had a working winbind configuration, in which
case the tests would just fail.)
Also, fix up ntlm-test a bit, so that if ntlm-test-helper for some
reason tries to authenticate as someone other than alice or bob, it
recognizes this, rather than considering everyone who isn't alice to
be bob.
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Move the auth-managery parts of SoupAuthManagerNTLM down into
SoupAuthManager, and the NTLMy parts into SoupAuthNTLM (in preparation
for supporting other kinds of connection-based auth such as
Negotiate).
The reorganization also makes it possible to use SoupAuthNTLM to
implement a mock version of /usr/bin/ntlm_auth, so we can extend
ntlm-test to test both the built-in-NTLM and external-NTLM codepaths.
Doing this reveals that, AFAICT, the external codepath did not
previously actually work, because it mis-used
G_SPAWN_FILE_AND_ARGV_ZERO and so ended up passing incorrect arguments
to /usr/bin/ntlm_auth.
|
|
|
|
|
|
| |
Add a test that each of the 3 SoupSession types (plain, async, sync)
behaves as expected with soup_session_queue_message(),
soup_session_send_message(), and soup_session_cancel_message().
|
|
|
|
|
|
|
| |
Update the docs to reflect the changes in plain SoupSessions vs its
traditional subclasses. And while we're there, add a few new ones
(including making soup_session_queue_message() and
soup_session_send_message() work on plain SoupSession.
|
|
|
|
|
|
|
|
|
|
|
| |
SoupMessage isn't being deprecated, and mirroring its API onto
SoupRequestHTTP is just going to result in always having to add every
new API twice. Also, it turns out to be less useful than originally
expected anyway, since you end up having to cast between SoupRequest
and SoupRequestHTTP frequently anyway.
This reverts commit d7117329400e47d2187ed033099d921d555f8d71 and most of
commit 53c270d0e2868fa5ad48ce864f10a9486b11a071.
|
|
|
|
|
|
| |
gtk-doc expects "Deprecated:" notes in docs to be matched up with some
deprecation-guard #define. So even though we don't want people to use
it any more, add LIBSOUP_DISABLE_DEPRECATED back.
|
| |
|
| |
|
| |
|
|
|
|
|
|
| |
Move the SoupCache special handling from SoupRequestHTTP to
SoupSession (as another step towards making SoupCache not such a
magical special case hack that only works in certain situations).
|
|
|
|
|
|
| |
Add a test of basic cache functionality (finally!)
Still needs more tests...
|
| |
|
| |
|
| |
|
| |
|