It may not be obvious, but the moment unlock is called, the locker
instance may be destroyed.
See g_object_unref(), which calls toggle_refs_check_and_ref_or_deref().
It will check for toggle references while dropping the ref count from 2
to 1. It must decrement the ref count while holding the lock, but it
also must still unlock afterwards.
Note that the locker instance is on the object itself. Once we decrement
the ref count we give up our reference and another thread may race
against destroying the object. We thus must not touch object anymore.
How can we then still unlock?
This works correctly because:
- unlock operations must not touch the locker instance after unlocking.
- assume that another thread races g_object_unref() to destroy the
object, while we are about to call object_bit_unlock() in
toggle_refs_check_and_ref_or_deref(). Then that other thread will also
need to acquire the same lock (during g_object_notify_queue_freeze()).
It thus is blocked to destroy the object.
Add code comments about that.
We can only assert for having one toggle reference, after we confirmed
(under lock) that the ref count was in the toggle case.
Otherwise, if another thread refs/unrefs the object, we can hit a wrong
g_critical() assertion about
if (tstackptr->n_toggle_refs != 1)
{
g_critical ("Unexpected number of toggle-refs. g_object_add_toggle_ref() must be paired with g_object_remove_toggle_ref()");
Fixes: 9ae43169cf ('gobject: fix race in toggle ref during g_object_ref()')
We don't actually need to use the Meson-detected size macros here,
because the result of `sizeof()` is an integer constant expression.
No functional change.
Signed-off-by: Simon McVittie <smcv@collabora.com>
g-ir-scanner currently maps these to lower-level types at scan time by
assuming that time_t is an alias for long, off_t is an alias for size_t
and so on. This is not always accurate: some ILP32 architectures have
64-bit time_t (for Y2038 compatibility) and 64-bit off_t (for large file
support), and that mismatch is tracked as GNOME/gobject-introspection#494.
One option for resolving this g-ir-scanner bug is to have it pass these
types through to the GIR XML, and teach g-ir-compiler and its replacement
gi-compile-repository to convert them to the corresponding concrete
type tag, as they already do for abstract types such as `long long` and
`size_t`.
Loosely based on GNOME/gobject-introspection!451 by Shuyu Liu.
Co-authored-by: Shuyu Liu <liushuyu011@gmail.com>
Signed-off-by: Simon McVittie <smcv@collabora.com>
We don't actually need to use the results of configure-time checks here:
sizeof is a perfectly reasonable integer constant expression, so we can
use that directly.
Helps: https://gitlab.gnome.org/GNOME/glib/-/issues/2842
Signed-off-by: Simon McVittie <smcv@collabora.com>
These scripts use $(readlink -f) to guess their own path if necessary,
but macOS readlink doesn't support the -f option, and POSIX doesn't
guarantee that readlink even exists.
Resolves: https://gitlab.gnome.org/GNOME/glib/-/issues/3289
Fixes: d7601f7e "Incorporate some lint checks into `meson test`"
Signed-off-by: Simon McVittie <smcv@collabora.com>
The `gi-docgen` tool is not designed to be used like that. In
particular, when nesting documentation directories, the generated
`*.devhelp2` files (needed by Devhelp to show the documentation) are
nested one directory level too deep for Devhelp to find them, and hence
are useless, and the documentation doesn’t show up in this common
documentation viewer.
So, change the installed documentation directory hierarchy:
* `${PREFIX}/share/doc/glib-2.0/gio` → `${PREFIX}/share/doc/gio-2.0`
* `${PREFIX}/share/doc/glib-2.0/glib-unix` →
`${PREFIX}/share/doc/glib-unix-2.0`
* `${PREFIX}/share/doc/glib-2.0/gobject` →
`${PREFIX}/share/doc/gobject-2.0`
* etc.
* `${PREFIX}/share/doc/glib-2.0/glib` → `${PREFIX}/share/doc/glib-2.0`
This is going to seem like pointless churn (the contents of the
documentation have not changed), and packagers may mourn the split of
content in `/usr/share/doc` from `/usr/share/doc/${package_name}` to
`/usr/share/doc/${pkg_config_id}` instead, but that seems to be the best
approach to fix this issue in GLib. gi-docgen’s behaviour does feel
fairly consistent and correct with the rest of how it works (single
output directory).
Signed-off-by: Philip Withnall <pwithnall@gnome.org>
Fixes: #3287
These unfortunately have `gchar*` return types rather than `const
gchar*`. This is a historical artifact which we can’t change: while
adding `const` would only be an API break and not an ABI break, it would
cause all sorts of C++ code which uses GLib to emit new cast warnings
(similarly, C code with const correctness compiler warnings enabled
would do the same).
The incorrect return type causes the GIR scanner to (reasonably) assume
the return value is allocated, which is wrong.
Fix that by explicitly adding `(transfer none)`.
Also add an explicit `(nullable)` because all three functions are.
Signed-off-by: Philip Withnall <pwithnall@gnome.org>
Fixes: #3286
The gdbus-example-objectmanager visibility header was being re-created
on reconfigure, causing a needless rebuild of gdbus tests that were
using the visibility header.
All other invocations of gen_visibility_macros are via custom_target.
The documentation previously implied that they could. That’s not really
true though: they can only fail if preconditions fail, i.e. they’re
passed invalid input. That’s a programmer error, which is not something
we want to encourage people to check for at runtime (e.g. by dynamically
checking for a 0 return value).
Signed-off-by: Philip Withnall <pwithnall@gnome.org>
This is nowhere near a complete check-through and gi-docgenification of
the signals docs, just a few bits I was looking at anyway.
Signed-off-by: Philip Withnall <pwithnall@gnome.org>
Helps: #3250
We install gvariant-specification-1.0.html to `docs_dir` but that shouldn't require `enable_gir`.
Fixes: docs/reference/glib/meson.build:26:16: ERROR: Unknown variable "docs_dir"
Part-of: <https://gitlab.gnome.org/GNOME/glib/-/merge_requests/3955>
