GitLab

Part-of: <!45>

Part-of: <!43>

see: #28

Part-of: <!42>

Thibault Saunier authored Aug 20, 2020 and GStreamer Marge Bot committed Aug 24, 2020

Part-of: <!40>

When introducing zero copy buffers/memory mapping we broke the API,
this brings back the exact same API as before for all the previously
handled cases but still raises an exception when using a context
to map buffers.

Fixes #40

Part-of: <!39>

Part-of: <!35>

- unref PyObject on exit with Py_XDECREF
- better error handling
- free memory view in case of erros

Part-of: <!34>

pyg_boxed_new cause a memory leak if it hold a copy of the boxed wrapper and freed when the wrapper is deallocated.
use the boxed wrapper value itself and don't hold a copy of the value.

pylib_loc = python.get_variable('LIBPL', '') alreay returns
the correct path for python's library dir

When returning those values, extensions must take a new reference.

Fixes #33

It's no longer used.

Olivier Crête authored May 21, 2017 and Guillaume Desmottes committed Jan 07, 2020

Fix #8

Since Python 3.8 the normal checks don't include the Python libraries
anymore and linking of the gst-python module would fail.

See also https://github.com/mesonbuild/meson/issues/5629

Fixes #28

With the new mapping API we can efficiently use the ->fill vmethod
which is sensibly better.

And minor cleanup in the way errors are handled

Daniel Klamt authored Apr 05, 2019 and Philipp Zabel committed Dec 09, 2019

The Problem is, that in the current state it is not easily possible to
edit the buffer data in a gstreamer python element since you get a copy
of the real buffer.

This patch overrides the mapinfo and the function generating it in a way
so that mapinfo.data is now a memoryview pointing to the real buffer.
Depending on the flags given for this buffer the memoryview is r/w.

We have notified application developers this would happen a long time
ago and python2 is going to be deprecated very soon now, before 1.18
is going to be released.

The previous commit broke those by trying to pass weak refs
through pygobject, but we should probably have tested the elements
beyond instantiation: weakref.WeakMethod returns a callable, but
that callable when called only returns the ephemeral bound method,
which is the object we want to call, but pygobject has no support
for that.

Instead, fix the memory leaks we were going after by decoupling the
lifecycle of the callback and that of the pad, by passing functors
to pygobject.

Mathieu Duponchelle authored Aug 09, 2019 and Mathieu Duponchelle committed Aug 08, 2019

Use weakref to avoid leaks, and remove refcount hack as the actual
issue has been fixed in pygobject

And add a test

See https://gitlab.gnome.org/GNOME/pygobject/issues/325