GTK lost it's '+' suffix back in 2019, according to
<https://mail.gnome.org/archives/gtk-devel-list/2019-February/msg00000.html>
This commit can be re-generated with:
git grep -l GTK+ \
| grep -v -e ^NEWS -e ^glib/tests/collate.c \
| xargs sed -i 's/GTK+/GTK/g'
Most of the changes are in comments and documentation.
Use more modern styling to the code added in the previous patch:
- split 'label: stmt; stmt;' into multiple lines
- add default: label with g_assert_not_reached() [yes, it's a bit
weird adding an assertion inside code that handles assertions, but
we should be okay since g_assertion_message_* are not public
functions and should only be used by our macros]
- use <inttypes.h> for shorter format strings
Note, however, that using uint64_t in gtestutils.h is not feasible,
since it would require adding an '#include <stdint.h>' with potential
unintended namespace pollution to older clients.
Signed-off-by: Eric Blake <eblake@redhat.com>
While x86_64 has enough precision in long double to do a round trip
from guint64 to long double and back, this is platform-specific, and
is a disservice to users trying to debug failing unit tests on other
architectures where it loses precision for g_assert_cmp{int,uint,hex}.
See also https://bugzilla.gnome.org/show_bug.cgi?id=788385 which
mentions having to add casts to specifically silence the compiler on
platforms where the precision loss occurs.
Meanwhile, g_assert_cmpuint() does an unsigned comparison, but outputs
signed values if the comparison fails, which is confusing.
Fix both issues by introducing a new g_assertion_message_cmpint()
function with a new 'u' numtype. For backwards compatibility, the
macros still call into the older g_assertion_message_cmpnum() when not
targetting 2.78, and that function still works when passed 'i' and 'x'
types even though code compiled for 2.78 and later will never invoke
it with numtype anything other than 'f'. Note that g_assert_cmpmem
can also take advantage of the new code, even though in practice,
comparison between two size_t values representing array lengths that
can actually be compiled is unlikely to have ever hit the precision
loss. The macros in signals.c test code does not have to worry about
versioning, since it is not part of the glib library proper.
Closes#2997
Signed-off-by: Eric Blake <eblake@redhat.com>
This should hopefully stop the kernel spending a lot of memory and disk
bandwidth creating coredumps for them unnecessarily, which slows down
the rest of the tests and generally wastes resources.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Fixes: #2939
This is put together through git archaeology:
```
git log -- glib/tests/assert-msg-test.c tests/assert-msg-test.c
```
The following commits were too trivial to have meaningful copyright:
- 8e59d8602ca5921d78245f5d2b405b517a5e7cf9
- 44c004c84e9080040ff4e0e90b322dda0561cf85
- 207b8cb8a50d68e207d28b59e588311a5cbd9772
- a1bee97d4f093db01dee834bf3292eabd5b13d17
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #1415
Apparently 3 hashes is too many, and as a result this line is not
displayed properly in the HTML documentation. 2 hashes seems to be
just right, as can be seen with:
$ git grep '^ \* ##' | wc -l
65
$ git grep '^ \* ###' | wc -l
1
Otherwise it might look like it would only start spawning threads when
jobs are enqueued.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #2958
If you’re only quickly looking at the API signature, it looks like
`item_free_func` will be called for all items enqueued to the thread
pool.
As it happens, it’s actually only called for the items which are still
enqueued when the thread pool is destroyed. The user’s `GFunc` is
responsible for freeing items which are successfully dequeued and
processed during the lifetime of the thread pool.
That’s a bit of a gotcha, so document it more explicitly.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
If both __NR_futex and __NR_futex_time64 are defined, g_futex_simple()
will first call futex_time64(). If that fails with ENOSYS, then
futex_time() is called instead. However, errno was not saved and
restored in this case, which would result in g_futex_simple()
returning with errno set to ENOSYS, even if futex_time() succeeded.
As with the previous commit.
The logic has to be a little contorted here to avoid leaving the context
locked after emitting the critical warning. Execution does (and should)
continue after a critical warning by default, so we should do our best
to recover.
Inspired by https://gitlab.gnome.org/GNOME/glib/-/merge_requests/3302.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
The documentation was fairly clear before, but we can make it clearer:
it’s a programming error to call `g_main_context_release()` if you have
not received a true return value from `g_main_context_acquire()` before.
Inspired by https://gitlab.gnome.org/GNOME/glib/-/merge_requests/3302.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
We need a way to initialise refcounted types placed in static storage,
or on the stack. Using proper macros avoids knowing the magic constant
used for grefcount and gatomicrefcount.
In non-perf mode, we were making only one thread to win the race to increase
the value, but since we're launching more threads anyways let's allow
more of them to have a chance to do something, to make the test more
valuable, even if it's still quick enough.
This removes the `comment` member of the GKeyFileGroup structure, which
seemed intended to distinguish comments just above a group from comments
above them, separated by one or more blank lines. Indeed:
* This does not seem to match any specification in the documentation,
where blank lines and lines starting with `#` are indiscriminately
considered comments. In particular, no distinction is made between the
comment above the first group and the comment at the beginning of the
file.
* This distinction was only half implemented, resulting in confusion
between comment above a group and comment at the end of the preceding
group.
Instead, the same logic is used for groups as for keys: the comment
above a group is simply the sequence of key-value pairs of the preceding
group where the key is null, starting from the bottom.
The addition of a blank line above groups when writing, involved in
bugs #104 and #2927, is kept, but:
* It is now added as a comment as soon as the group is added (since a
blank line is considered a comment), so that
`g_key_file_get_comment()` returns the correct result right away.
* It is always added if comments are not kept.
* Otherwise it is only added if the group is newly created (not present
in the original data), in order to really keep comments (existing and
not existing).
Closes: #104, #2927
Even though having NULL as nullptr should be the standard for newer C++
versions, it may break some headers, so let's not touch it for now.
Closes: https://gitlab.gnome.org/GNOME/glib/-/issues/2973
A context iteration we're doing lots of lock/unlocks and that's fine to give
other threads contexts a chance to run, but we're doing it also just to call
other functions that require locking, and this can be avoided.
Other threads can still have a chance to run while releasing the ownership
of the context.
In practice, this will never happen.
If `getcwd()` returns `ERANGE` whenever the working directory is ≥
`PATH_MAX`, though, the previous implementation of the loop would run
until `max_len == G_MAXULONG`, and would then overflow when adding `1`
to it for a nul terminator in the allocation.
Avoid that problem by always keeping `buffer_len` as a power of two, and
relying on `getcwd()` to write a nul terminator within the buffer space
it’s given. It seems to do this on all platforms we care about, because
the previous version of the code never explicitly wrote a nul terminator
anywhere.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Fixes: #98
Various projects are running tests under valgrind, and they are using
the GLib suppresions to avoid false-positive results.
While this is stored in a well-known path for some years, and easy to
figure out from the GLib prefix, it's better to expose it through a
proper pkgconfig variable so that it's easy to get it from any build
system.
We were working around that with a call to chdir("."). Internally chdir()
sets up cmd shell-related environment variables, but only if the current
working directory belongs to a volume with a drive letter. For example,
if the current working directory is on a UNC volume like a network share,
the call to chdir() won't help.
Reference:
https://developercommunity.visualstudio.com/t/UCRT-Crash-in-_wspawne-functions/10262748
Add three classes of test case for which g_utf8_normalize() should
safely return NULL:
- Strings ending with a truncated multibyte character which would
extend past the NUL terminator
- Strings ending with a multibyte character which extends past the
length limit provided by the max_len argument
- Strings containing an invalid multibyte character in any position
_g_utf8_normalize_wc() could read past the end of the provided buffer if
it ends with a truncated multibyte character. If max_len is -1, it can
continue reading until it encounters either a NUL or unreadable
memory. Avoid this with extra bounds checks prior to g_utf8_get_char()
to ensure that it does not read past either max_len or a NUL
terminator.
