glib/gobject
Philip Withnall e41b4b1acb
gclosure: Split out invalidation to a helper function
This avoids the need to ref/unref the closure while invalidating it in
the `closure->ref_count == 1` path in `g_closure_unref()`.

scan-build gets very confused about the ref count here, and ends up
assuming it’s possible for the `g_closure_unref()` call in
`g_closure_invalidate()` to finalise the closure when the latter is
called from `g_closure_unref()`. There was an existing assertion in
`g_closure_invalidate()` which hinted that this wasn’t possible, but
scan-build doesn’t seem to be able to propagate assumptions about
refcounts between function contexts.

So, introduce an internal variant of `g_closure_invalidate()` which can
skip modifying the closure’s refcount. It’s safe to invalidate the
closure without adding a ref when doing so from `g_closure_unref()` with
`closure->ref_count == 1` because at that point `g_closure_unref()`
holds the only remaining ref to the closure. So none of the invalidation
callbacks are allowed to unref it further.

Signed-off-by: Philip Withnall <pwithnall@gnome.org>

Helps: #1767
2024-04-25 23:57:27 +01:00
..
2023-10-16 23:34:04 +01:00
2023-12-31 09:09:48 +00:00
2022-10-13 20:53:56 -04:00
2023-10-16 23:35:05 +01:00
2023-12-31 09:09:48 +00:00
2022-10-13 20:53:56 -04:00
2023-10-16 23:34:24 +01:00
2024-04-01 11:01:06 +00:00
2023-12-19 18:41:02 +00:00
2022-10-13 20:53:56 -04:00
2022-10-13 20:53:56 -04:00
2024-04-01 11:01:06 +00:00
2023-10-23 11:26:53 +01:00
2022-10-13 20:53:56 -04:00
2024-04-01 11:01:06 +00:00