| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
| |
Our plugin raising exceptions seems to confuse other hooks, marking it
to be run first gets rid of them.
|
| |
|
| |
|
|
|
|
|
| |
default_timer() gives us the most precise time information on all platforms.
time.clock() will be removed with Python 3.8
|
|
|
|
| |
It will stop to work with Python 3.8
|
| |
|
|
|
|
| |
Allow failures for now as we don't control it and can't easily react if it breaks.
|
|
|
|
|
|
|
|
| |
It imports the modules directly and puts them in sys.modules under the full
name. gi._gi_cairo was missing which made it import the system version
(because the gi package is registered as a namespace package).
Handle gi._gi_cairo there as well.
|
| |
|
|
|
|
| |
Noticed because the gnome sdk image has pycairo 1.14.1
|
|
|
|
|
| |
msys2 has updated to gcc8 and lcov can't read the resulting files.
Maybe this helps.
|
|
|
|
|
| |
This sometimes segfaults on pypy3. I can't see why, so skip for now
and reenable once pypy3 haa a stable release.
|
|\
| |
| |
| |
| | |
glib.wrap: Update urls
See merge request GNOME/pygobject!80
|
|/
|
| |
No idea for what this is used, but the urls needs update.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
in an overlay. Fixes #230
g_resources_get_info() doesn't do overlays but we use it to verify that the resource
exists. See https://gitlab.gnome.org/GNOME/glib/issues/1445
In case g_resources_get_info() errors out check g_resources_lookup_data() before giving up.
Because glib reads the overlay paths only once at start this is hard to add tests for.
|
|\
| |
| |
| |
| | |
pygi-convert: Fix some issues
See merge request GNOME/pygobject!78
|
| | |
|
| | |
|
| |
| |
| |
| | |
theoretically gitlab bans should be disabled now
|
| |
| |
| |
| |
| |
| | |
The attribute deletion here is the easiest way to stop current PyGObject
from double-handling the initialisation in this case, which breaks the
newer init_template() handling.
|
|/
|
|
|
|
|
|
| |
Without this change, the second instantiation of any templated class
will fail to run any of the actual template-handling logic.
This is *only* relevant when using the _gtk_template helpers with an
older PyGObject; in newer versions, this method isn't called.
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
| |
Match what we do with distutils
* Use "PyGObject" as package name (it's not case sensitive, but no need to
diverge there)
* Mark unstable releases as dev releases by adjusting the version number
* Install in the right directory (not the gi package)
|
| |
|
|
|
|
|
|
|
|
|
| |
new_from_data isn't bindable (see https://bugzilla.gnome.org/show_bug.cgi?id=721497)
and will result in use after free. While we could try to move people away from it and
skip it in GI a github code search turns up quite a few users.
This adds an override for it, wrapping GdkPixbuf.Pixbuf.new_from_bytes, and marks any
usage of the callback args as deprecated.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Using a GPtrArray as the value type of a GHashTable currently produces a
warning:
Unsupported type array
for example when calling a function whose return type is annotated as:
(transfer container) (element-type filename GPtrArray<SomeObject>)
Since we don’t treat objects (GI_TYPE_TAG_INTERFACE) specially when
marshalling to/from GHashTable, we shouldn’t need to treat arrays
specially either. Handle the case explicitly.
Signed-off-by: Philip Withnall <withnall@endlessm.com>
|
|
|
|
|
|
| |
pacman doesn't support installing packages without doing a system upgrade
first. In case the user already has msys2 installed but not used in a while
just doing the install might not work.
|
| |
|
| |
|
|
|
|
| |
we run the whole test suite as one test
|
|
|
|
| |
pypy+pycairo doesn't work because it doesn't install a .pc file with it.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
through SIGINT. Fixes #219
We didn't remove them if the event loops returned normally.
This also fixes a small race where a SIGINT gets ignored right after the yield check and before
the old handler is reinstated. Check after the signal handler is switched back instead.
This resulted in GLib.MainLoop and Gio.Aplication instances leaking.
|
| |
|
|
|
|
|
|
| |
disable it
If you don't want pycairo support pass "-Dpycairo=false"
|
| |
|
|\
| |
| |
| |
| | |
docs: Add introduction to handling GLib.Error
See merge request GNOME/pygobject!69
|
| | |
|
| | |
|
| | |
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
After chaining up accessing anything on the instance will crash like
accessing self.g_type in this case. Guard against that by exposing if the
wrapped boxed is still there.
Ideally we shouldn't invalidate the object in __del__, but various
things depend on it atm. For another time..
|
| |
| |
| |
| |
| | |
Debian seems to have started using LTO for Python and some symbol names
have a ".lto_priv" suffix now.
|
|/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
If a GValueArray is created from a list of GValues, it should not wrap
these GValues in a second layer of GValues before appending to the
array. The end result should be a GValueArray of GValues, not a
GValueArray of GValues holding another GValue (unless the user
explicitly creates such a value).
With this patch the behavior is now consistent when creating a
GValueArray and appending GValues directly, and creating a GValueArray
within a GValue based on a list of GValues.
For instance, to create a GValueArray of G_TYPE_UINT the user must
create GValues manually and create a GValueArray from these. The result
should be a GValueArray of GValues that hold G_TYPE_UINT, not a
GValueArray that contain GValues that hold a GValue that holds
G_TYPE_UINT.
See !66
|
|
|
|
|
|
|
|
|
| |
find_module() should either return None or a loader, but we raised ImportError
there in case we already knew that the namespace was missing.
Move that check to load_module() instead. While there shouldn't be any functional
difference, raising in find_module() under Python 3 resulted in a chained
exception with an unrelated error message printed first.
|