framework
* configure.ac:
* tests/*:
The tests have been reorganised into different categories: conformance,
interactive and micro benchmarks.
- conformance tests can be run as part of automated tests
- interactive tests are basically all the existing tests
- micro benchmarks focus on a single performance metric
I converted the timeline tests to conformance tests and also added some
tests from Neil Roberts and Ebassi.
Note: currently only the conformance tests use the glib test APIs,
though the micro benchmarks should too.
The other change is to make the unit tests link into monolithic binaries
which makes the build time for unit tests considerably faster. To deal
with the extra complexity this adds to debugging individual tests I
have added some sugar to the makefiles so all the tests can be run
directly via a symlink and when an individual test is run this way,
then a note is printed to the terminal explaining exactly how that test
may be debugged using GDB.
There is a convenience make rule: 'make test-report', that will run all
the conformance tests and hopefully even open the results in your web
browser. It skips some of the slower timeline tests, but you can run
those using 'make full-report'
* clutter/clutter-model-default.[ch]: Rename ClutterModelDefault
into ClutterListModel, to make it clear that the default
ClutterModel implementation is list-based
* tests/test-model.c: Update call
* clutter/clutter-model-default.c:
(clutter_model_default_iter_next),
(clutter_model_default_iter_prev): Do not instantiate a new
iterator, and update the passed one instead, as we say we do
in the documentation. This avoid leaking tons of iterators.
* clutter/clutter-model.[ch]: Update the documentation.
* tests/test-model.c: Prettify some output.
* clutter/clutter-behaviour.c:
(clutter_knot_get_type): Use the I_() macro.
* clutter/clutter-model.[ch]: Make ClutterModel and ClutterModelIter
abstract classes; also, provide more virtual functions inside the
ClutterModel vtable, to allow subclasses and language bindings to
override the entire ClutterModel using other/native data types
to store the rows.
* clutter/clutter-model-default.[ch]: ClutterModelDefault is a
default implementation of ClutterModel and ClutterModelIter using
GSequence.
* clutter/Makefile.am:
* clutter/clutter.h: Build glue for ClutterModelDefault.
* tests/test-model.c: Update for constructor changes.
* tests/test-shader.c: Fix a typo.
* clutter/clutter-model.[ch]: Slight API change in the constructor
functions for ClutterModel: clutter_model_new() now takes a list
of GType/string pairs for both the column type and the column name.
(clutter_model_set_n_columns),
(clutter_model_set_names),
(clutter_model_set_types): Subclasses of ClutterModel will be able
to call clutter_model_set_types() and clutter_model_set_names() in
any order, provided that they don't call each function more than
once.
* tests/test-model.c: Update the test case.