This bug only gets triggered on BSD, because it calls `recv_message()`
with `G_IO_HUP | G_IO_IN`. That takes two code paths in
`recv_message()`, and ends up calling `shutdown_source()` twice. The
second call causes a critical warning.
Fixes: https://gitlab.gnome.org/GNOME/glib/-/jobs/2004192
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
The g_spawn_async_with_pipes_and_fds() API changes its behavior
depending on the states of out parameters. Apart from the philosophical
inconsistency, this poses some real problems, e.g. in Lua bindings:
https://github.com/lgi-devs/lgi/issues/277
This adds three new spawn flags to allow specifying the wanted behavior
explicitly and does not assert on the state of the output parameters.
The runner machine is offline and there is no ETA on when it will be
back, so disable the CI job which uses it for now so that pipelines
can proceed.
See https://gitlab.gnome.org/Infrastructure/GitLab/-/issues/558
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Check the spawn implementation behaviour when the stderr is a
socket (mostly for win32).
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Used in following commits, including in some GIO experiments, so make it
a private API.
For now, this implementation is similar to the glib/gspawn-win32.c one,
with mroe error checking and better on error behaviour. A following
patch will also fix the case of duplicating sockets.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Use a prefix matching the binary, rename the test "scm", as it involves
SocketControlMessage.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
This improves the formatting, but doesn’t touch the content of the file,
some of which is quite out of date.
I’ll approach rewriting the file in a later commit/MR.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
We hadn’t been substituting the version number in properly for a while,
so maybe just abandon that approach and rewrite the instructions to not
refer to a specific version number.
This simplifies things a bit, and means the install instructions can be
read online in GitLab.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
This is part of an effort to rationalise our developer documentation a
bit, putting it all in one directory rather than cluttering up the root
directory.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
It was basically a human-readable symlink.
`README.md` seems widely enough accepted now that tooling hopefully
won’t break if we delete the non-Markdown version.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
There’s no point in having it in the root directory when it’s only
really used for CI. It just clutters up the root.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
The git log (which didn’t exist when this file was written), and
copyright statements, give credit to these original developers.
Listing their names and e-mail addresses so prominently seems like a
recipe for random people to try e-mailing them for support.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
This encodes what GLib is already bound by, as it’s hosted on
gitlab.gnome.org, which is GNOME infrastructure and hence bound by the
GNOME code of conduct.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
