The two test scripts actually assumed some *NIX paradigms, so we need
to adapt them so that they can work on Windows as well, the changes are
namely:
-Call the glib-mkenums and glib-genmarshal Python scripts with the
Python interpreter, not just relying on shebang lines, on Windows.
This is because the native Windows console (cmd.exe) does not support
shebang lines, for subprocess.run().
-Use NamedTemporaryFile with delete=False, otherwise Windows cannot find
the temp files we need when running the tests.
-Use universal_newlines=True for subprocess.run() so that we do not need
to worry out line ending differences on different systems.
-Make sure we are not in the temp directories we create, where the tests
are being run, upon cleanup. Windows does not like deleting
directories that we are currently in.
When building a valist marshaller, we can avoid reffing a GParamSpec
if the argument is known to always be static. The marshaller we ship in
`gmarshal.c` got this right, but marshallers generated by
glib-genmarshal were missing the optimisation. Fix that, and add a unit
test.
Signed-off-by: Philip Withnall <withnall@endlessm.com>
Fixes: #1792
When building a valist marshaller, we can avoid a string copy if the
argument is known to always be static. The marshaller we ship in
`gmarshal.c` got this right, but marshallers generated by
glib-genmarshal were missing the optimisation. Fix that, and add a unit
test.
Signed-off-by: Philip Withnall <withnall@endlessm.com>
Fixes: #1792
The old (Perl) implementation of glib-genmarshal used
g_variant_ref_sink() to correctly handle floating inputs; the Python
version should do the same.
Includes a unit test.
Signed-off-by: Philip Withnall <withnall@endlessm.com>
Fixes: #1793
This is a basic test suite for the `glib-genmarshal` utility, lifted
mostly directly from the tests for `glib-mkenums`.
Signed-off-by: Philip Withnall <withnall@endlessm.com>
