* Added grl_init
* Replaced KeyID constant with a variable looked up
* Tested search callback function

As gjs currently doesn't support GParamSpecs (see #626047), this
patch works around it by replacing them with uints.

Damien Lespiau authored Nov 12, 2010 and Victor Manuel Jaquez Leal committed Nov 12, 2010

Remove the const before the GList *, shutting down the warnings,
make the global variables static.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>

Damien Lespiau authored Nov 12, 2010 and Victor Manuel Jaquez Leal committed Nov 12, 2010

write (2) needs the inclusion of unistd.h.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

Damien Lespiau authored Nov 02, 2010 and Victor Manuel Jaquez Leal committed Nov 02, 2010

Using $(top_builddir)/libs/net breaks automake's dependency between
wc-test and libgrlnet. When using recursive builds, Automake
automatically tracks the dependencies in the same directory but when
using $(top_builddir)/libs/net/libgrlnet.la it assumes the library is in
another directoy and is already build (by assuming the other directory is
before the current one in the parent's SUBDIRS).

By removing the $(top_builddir)/libs/net in wc_test_LDADD Automake now
understands that the library is in the same directory and it should be
built before wc-test and adds the dependency which fixes parallel
builds.

Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>

Simon Pena authored Oct 04, 2010 and Victor Manuel Jaquez Leal committed Oct 22, 2010

Added a configuration setup for Flickr, Youtube and Vimeo

Signed-off-by: Simón Pena <spenap@gmail.com>

Simon Pena authored Oct 04, 2010 and Victor Manuel Jaquez Leal committed Oct 22, 2010

* grl_config_new now allows NULL sources for a plugin-wide
  configuration
* grl_config_new is annotated as 'transfer none', to avoid freeing
  the configuration while the plugin is valid.

Signed-off-by: Simón Pena <spenap@gmail.com>

It makes more sense. I would like to name it something like
intersect_key_list() but the function parameters doesn't follow a logical
structure for that, so I adhered to the filter semantic.

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

When the source doesn't have slow/writable keys (empty set) the key list is
not filtered. Also, that code is repeated in each filter type.

This patch set to NULL the filtered key list, and the returned complement is
the passed key list. The code is isolated in the generic_filter() too.

Finally, added a description of the function.

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

Simon Pena authored Oct 04, 2010 and Victor Manuel Jaquez Leal committed Oct 20, 2010

Updated GrlMetadataSource filter tests so that they assert that
the aggregated length of the two lists returned (the one which
passes the check and the one which doesn't) equals to the original
key list.

Signed-off-by: Simón Pena <spenap@gmail.com>

Simon Pena authored Oct 04, 2010 and Victor Manuel Jaquez Leal committed Oct 20, 2010

grl_metadata_source_filter_{supported, slow, writable} methods
have a common filtering part and some specific bits each one.

This patch extracts the common code into a generic function so
the other ones can do their checks and specific behavior, and then
do the filtering.

Signed-off-by: Simón Pena <spenap@gmail.com>

If grilo-net dependencies are not found, it were still trying to build
grilo-net.

This patch fixes this problem.

Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

Added a wrong const identifier in callback parameter

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

This is one of the risks to be living on the edge.

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>

Signed-off-by: Iago Toral Quiroga <itoral@igalia.com>

Input arguments don't need 'transfer none' annotations.

Signed-off-by: Simón Pena <spenap@gmail.com>
Signed-off-by: Iago Toral Quiroga <itoral@igalia.com>

Simon Pena authored Sep 29, 2010 and Victor Manuel Jaquez Leal committed Sep 30, 2010

setUp/tearDown methods in the tests load and unload the plugins
so that each test is run in a clean environment.

However, the current plugin mechanism doesn't allow that, and reloading
a previously loaded/unloaded plugin produces a crash. This patch
removes setUp/tearDown code, adding the plugin load in the tests constructor
but leaving the plugins loaded during all the test suite execution. While
this is a fix which allows the tests to be run, some of them will fail.

Once that the plugin system is reworked to allow unloading/reloading, this
commit should be reverted to come back to the original behavior.

Simon Pena authored Sep 29, 2010 and Victor Manuel Jaquez Leal committed Sep 30, 2010

This patch tests the class' methods from Python using
GObject Introspection, checking functionality and bindings
at the same time.

Simon Pena authored Sep 29, 2010 and Victor Manuel Jaquez Leal committed Sep 30, 2010

Test runner script now accepts command line arguments. That way,
it can be used to check individual test suites, instead of running
the full test set. While this feature isn't really necessary, it
can be helpful when designing the tests.

Simon Pena authored Sep 29, 2010 and Victor Manuel Jaquez Leal committed Sep 30, 2010

    * Update configure.ac to generate a 'util' Python class
    * Added a test runner to launch Python unit tests
    * Added python/ to tests' SUBDIRS
    * Updated test's Makefile.am to avoid launching Python tests twice
      when 'make check' is invoked