Age | Commit message (Collapse) | Author | Files | Lines |
|
This fixes a compilation warning about a missing prototype.
|
|
|
|
|
|
https://bugzilla.gnome.org/show_bug.cgi?id=735922
|
|
https://bugzilla.gnome.org/show_bug.cgi?id=735921
|
|
https://bugzilla.gnome.org/show_bug.cgi?id=735921
|
|
https://bugzilla.gnome.org/show_bug.cgi?id=735921
|
|
The tests used to use term.ie which is gone.
There are still some intermittent failures when running the tests,
sometimes the test is successful, sometimes it's not :(
https://bugzilla.gnome.org/show_bug.cgi?id=735921
|
|
https://bugzilla.gnome.org/show_bug.cgi?id=735921
|
|
https://bugzilla.gnome.org/show_bug.cgi?id=735921
|
|
This is deprecated and causes compile-time warnings
https://bugzilla.gnome.org/show_bug.cgi?id=735921
|
|
This is deprecated and causes a compile warning
https://bugzilla.gnome.org/show_bug.cgi?id=735921
|
|
https must be used, and the correct URL for uploads is up.flickr.com
https://bugzilla.gnome.org/show_bug.cgi?id=735920
|
|
It was in a separate rest-proxy-call.h header which allowed derived
classes to access RestProxyCall private data without any control.
Better to go through explicit methods to do that.
https://bugzilla.gnome.org/show_bug.cgi?id=735919
|
|
Now that LastFMProxyCall no longer directly modifies RestProxyCall::url,
it's trivial to remove all accesses to RestProxyCallPrivate.
https://bugzilla.gnome.org/show_bug.cgi?id=735919
|
|
Now that FlickrProxyCall no longer directly modifies RestProxyCall::url,
it's trivial to remove all accesses to RestProxyCallPrivate.
https://bugzilla.gnome.org/show_bug.cgi?id=735919
|
|
OAuthProxyCall was directly accessing private RestProxyCall data. The
same functionality can be achieved using only librest public API, so
let's use that instead.
https://bugzilla.gnome.org/show_bug.cgi?id=735919
|
|
The oauth "prepare" step needs RestProxyCall::url to be set, but since
commit c66b6d this is only set after the "prepare" and
"serialize_headers" methods have been called. As it's
better if OAuthProxyCall do not directly access RestProxyCall private
data, this commit adds a getter to RestProxyCall which will be used
to make sure RestProxyCall::url is set when OAuthProxyCall::prepare
needs it.
https://bugzilla.gnome.org/show_bug.cgi?id=708359
|
|
Since commit c66b6d, RestProxyCall::url is regenerated after calling the
RestProxyCall::prepare virtual method. This breaks the lastfm code as
it needs to reset RestProxyCall::url in order to remove the function
from it.
It used to do that by directly accessing RestProxyCall private
data. Since c66b6d, this can be solved using only public methods as
if the function is reset on the RestProxyCall in ::prepare, ::url will
be regenerated to reflect that after ::prepare and ::serialize_params
have been called.
https://bugzilla.gnome.org/show_bug.cgi?id=708359
|
|
Since commit c66b6d, RestProxyCall::url is regenerated after calling the
RestProxyCall::prepare virtual method. This breaks the flickr code as
it needs to reset RestProxyCall::url in order to remove the function
from it.
It used to do that by directly accessing RestProxyCall private
data. Since c66b6d, this can be solved using only public methods as
if the function is reset on the RestProxyCall in ::prepare, ::url will
be regenerated to reflect that after ::prepare and ::serialize_params
have been called.
https://bugzilla.gnome.org/show_bug.cgi?id=708359
|
|
|
|
Since the public API now references a type from libsoup, we need
to add Soup-2.4.gir to Rest_@API_VERSION_AM@_gir_INCLUDES.
https://bugzilla.gnome.org/show_bug.cgi?id=728340
|
|
This function can be helpful if one wants more control over libsoup
features than what librest simple API provide. For example, to get full
access to libsoup cookie API (say to be able to add arbitrary cookies to
the soup session), one can do:
RestProxy *proxy = g_object_new(REST_TYPE_PROXY,
"url-format", url,
"disable-cookies", TRUE,
NULL);
SoupSessionFeature *cookie_jar = SOUP_SESSION_FEATURE(soup_cookie_jar_new ());
rest_proxy_add_soup_feature(proxy, cookie_jar);
It's then possible to use all the soup_cookie_* methods to deal with
cookies.
https://bugzilla.gnome.org/show_bug.cgi?id=728340
|
|
RestProxyCall::prepare_message() calls the serialize_params vfunc
so that child classes can serialize the call parameters in whichever
way they want.
One way of doing that could be to append the parameters to the URI
that is called (http://example.com?param1=value1;param2=value2).
However, the URI to call is determined at the beginning of
prepare_message(), and is not refreshed after calling
RestProxyCall::serialize_params(), so it's not possible to
append parameters to the URI that is going to be called.
This commit rebuilds the URI to call after calling serialize_params()
in case it has changed.
https://bugzilla.gnome.org/show_bug.cgi?id=708359
|
|
There are rest_proxy_call_[gs]et_method() methods, a
rest_proxy_call_set_function() method, but no corresponding getter.
Adding it makes the API more consistent.
https://bugzilla.gnome.org/show_bug.cgi?id=708358
|
|
|
|
When building the HTTP URL to access for a REST operation,
RestProxyCall::prepare_message, a '/' at the end of the base URL
is ignored, but a '/' at the beginning of the REST function is not.
When interacting with the oVirt REST API, I often end up building calls
with for example 'https://ovirt.example.com' as the base URI, and '/api/vms/'
as the function, which then leads to an URI with 2 '/':
'https://ovirt.example.com//api/vms'
https://bugzilla.gnome.org/show_bug.cgi?id=728952
|
|
rest_params_add() is currently using g_hash_table_insert() to add
the passed in parameter in the parameter hash table. The key which
is used is owned by the associated value.
When using rest_params_add to replace an already existing parameter,
the existing value will be freed with rest_param_unref().
However, g_hash_table_insert() does not replace the key when it already
exists in the hash table: "If the key already exists in the GHashTable
its current value is replaced with the new value... If you supplied a
key_destroy_func when creating the GHashTable, the passed key is freed
using that function."
This means that after replacing an already existing parameter, the
corresponding key will still be the old one, which is now pointing
at freed memory as the old value was freed.
g_hash_table_replace() ensures that the key will still be valid, even
when replacing existing parameters:
"Inserts a new key and value into a GHashTable similar to
g_hash_table_insert(). The difference is that if the key already exists
in the GHashTable, it gets replaced by the new key."
https://bugzilla.gnome.org/show_bug.cgi?id=665716
|
|
If given a URI it can't parse, soup_error_new() will return NULL.
https://bugzilla.gnome.org/show_bug.cgi?id=728953
|
|
|
|
|
|
|
|
|
|
It seems that http://term.ie/oauth/example/ is either down for the
count, or down for good. In the meantime, let's try to be resilient
against service outage at least during distcheck.
|
|
Both pause() and unpause() are referenced elsewhere in the API
reference, but have no formal documentation annotations.
https://bugzilla.gnome.org/show_bug.cgi?id=712747
|
|
https://bugzilla.gnome.org/show_bug.cgi?id=712747
|
|
The latter, especially, avoids gtk-doc complaining about introspection
annotations in the documentation comments.
https://bugzilla.gnome.org/show_bug.cgi?id=712747
|
|
It's an XML file, let's use the right extension.
https://bugzilla.gnome.org/show_bug.cgi?id=712747
|
|
https://bugzilla.gnome.org/show_bug.cgi?id=712747
|
|
https://bugzilla.gnome.org/show_bug.cgi?id=712747
|
|
So we have a fighting chance to actually fix them, instead of sweeping
them under the rug.
https://bugzilla.gnome.org/show_bug.cgi?id=712747
|
|
Syntax matters.
https://bugzilla.gnome.org/show_bug.cgi?id=712747
|
|
And this, kids, is why patches ought to be reviewed.
https://bugzilla.gnome.org/show_bug.cgi?id=712747
|
|
INCLUDES is deprecated and should not be used in modern autotools.
https://bugzilla.gnome.org/show_bug.cgi?id=712747
|
|
The SoupLogger instance needs to be attached last (at least after the
SoupCookieJar librest uses), otherwise REST_DEBUG=proxy will not dump the
cookie headers added by the SoupCookieJar instance.
https://bugzilla.gnome.org/show_bug.cgi?id=712231
|
|
The returned type was wrong as well as who owns the returned memory,
which could lead to memory corruption.
|
|
This property maps to the SoupSession::ssl-ca-file property, and
allows to specify a file that contains the CAs to use to verify
the certificates we'll get during a TLS session.
|
|
Based on a patch from Federico Mena Quintero for glib-networking
|
|
In the previous commit the downloading of binary data was fixed -
unfortunately this broke the behaviour of downloading textual data where
the clients expect a NUL terminated string for the payload.
This change makes sure to include the zero byte that soup adds to
support this use case and as a result the test suite passes.
|
|
This fixes an issue when a GSoC student was trying to fetch images from
gravatar using librest.
|