They make everything a lot harder to reason about, and easily allow for
state to leak between tests.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
It’s a tradeoff of time against reproducibility of the failure conditions
this test is testing for. If this test is run 100× on CI (which it will
be every few weeks), that should be often enough to catch a regression
here.
A regression in this code is unlikely, though.
This change is motivated by the fact that periodically this test times
out, and even when it doesn’t, it takes on average 240s of CI runner
time during each CI run. That’s a lot of resources.
See: https://gitlab.gnome.org/GNOME/glib/-/jobs/1862013
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
As well as with `EACCES`. This can happen if the filesystem as a whole
is read-only, and has been spotted in the wild on macOS:
```
Bail out! GLib:ERROR:../glib/tests/fileutils.c:756:test_mkdir_with_parents: assertion failed (errno == EACCES): (1 == 13)
stderr:
**
GLib:ERROR:../glib/tests/fileutils.c:756:test_mkdir_with_parents: assertion failed (errno == EACCES): (1 == 13)
```
See: https://stackoverflow.com/q/68766316/2931197
See: https://gitlab.gnome.org/GNOME/glib/-/jobs/1866500
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Manipulating PATH the way the test does, it breaks DLL lookups and fails
to run helper programs.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Add test-spawn-sleep helper program to sleep on Windows (child-test
removed in commit 241b9f41b, probably breaking the test)
(fwiw, Windows has a timeout command nowadays, but it conflicts with
msys2 timeout which has different usage)
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Test failures were previously ignored on macOS because there are 12
tests which consistently fail (and have not yet been fixed, because
there are no regularly active macOS maintainers for GLib; you could help
here!).
However, this means that new test failures can’t be spotted.
So, explicitly mark those 12 tests as `should_fail` on macOS, and then
make other test failures cause failure of the CI run.
We can track the process of fixing those 12 tests on #1392 and #1251.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #1392
The source needs to be removed from the `GMainContext` before being
unreffed, otherwise the main context and main loop will be kept around.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
`g_environ_setenv()` and `g_environ_unsetenv()` were correctly returning
`NULL` when their preconditions failed (as the test is supposed to be
exercising). That overwrote the value of `env` without freeing it,
resulting in a leak.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
We’re trying to eliminate the legacy `tests/` directory. This commit
moves the code from `tests/iochannel-test.c` into
`glib/tests/io-channel.c` and ports it to the latest GLib test coding
standards:
* Change `g_assert()` to `g_assert_*()`
* Print verbose messages with `g_test_message()`
* Rename some variables to conform to modern conventions
* Use `GTest`
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #1434
We can't exercise precondition check failures if GLib was (inadvisably)
compiled with -Dglib_checks=false, and we shouldn't necessarily exercise
precondition check failures when using QA tools like valgrind, so skip
these tests if run with -m no-undefined.
Signed-off-by: Simon McVittie <smcv@collabora.com>
Previously, these tests would always pass. If the precondition check
failed (as we want it to), the subprocess would exit unsuccessfully;
but if the precondition check wrongly passed, the subprocess would
continue, allocate a nonzero amount of memory, and fail the
g_assert_null(), resulting in the subprocess exiting unsuccessfully
and the test still passing.
Signed-off-by: Simon McVittie <smcv@collabora.com>
On ILP32 platforms, 4 is a valid alignment for g_aligned_alloc(), so
use 2 as our invalid alignment instead.
Signed-off-by: Simon McVittie <smcv@collabora.com>
This can happen if a caller (ab)uses `execve()` to execute a gtest
process with an empty `argv` array. `g_test_init()` has to gracefully
handle such a situation.
Fix a few problem areas in the code, and add a simple test which checks
that `g_test_init()` doesn’t crash when called with an empty `argv`.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
