Various tests have leaks where it isn't clear whether the data is
intentionally not freed, or leaked due to a bug. If we mark these
tests as TODO, we can skip them under AddressSanitizer and get the
rest to pass, giving us a baseline from which to avoid regressions.
Signed-off-by: Simon McVittie <smcv@collabora.com>
AddressSanitizer detects memory leaks, NULL parameters where only a
non-NULL parameter is expected, and other suspicious behaviour, so if
we try to test that sort of thing we can expect it to fail.
Signed-off-by: Simon McVittie <smcv@collabora.com>
AddressSanitizer, UndefinedBehaviourSanitizer and probably others
involve adding instrumentation into the code under test, which doesn't
go well with LD_PRELOAD modules that absolutely need to be
self-contained.
Signed-off-by: Simon McVittie <smcv@collabora.com>
We have a "good" implementation of g_clear_signal_handler() in
form of a macro. Use it, and don't duplicate the code.
Also add a comment to the documentation that "instance" in fact must
not point to a valid GObject instance -- if the handler ID is unset.
Also reword the documentation about the reasoning for why a macro
version exists. The reason is not to use the function "without
pointer cast". I don't think the non-macro version requires any
pointer cast, since "instance" is a void pointer. Was this referring
to the handler_id_ptr? That doesn't seem right either, because the
caller should always provide a "gulong *" pointer and nothing else.
Preferably macros behave function-like to minimize surprises. That
means for example that they evaluate all arguments exactly once.
Rework g_clear_signal_handler() to assign the macro parameters
to auto variables so they are accessed exactly once.
Also, drop the static assert for the size of (*handler_id_ptr).
As we now assign to a "gulong *" pointer, the compiler already
checks the types. In fact, the check is now stricter than before.
Previously it would have allowed a pointer to a "signed long".
This is a change in behavior of the macro and the stricter compile
check could cause a build failure with broken code.
Also, clear the handler id first, before calling
g_signal_handler_disconnect(). Disconnecting a signal invokes the
destroy notify, which can have side effects. It just feels cleaner
to first reset the *_handler_id_ptr, before those side effects
can happen. Of course, in practice it makes little difference.
g_signal_new_valist() is called by g_signal_new(), which is probably
the most common way to create a signal.
Also, in almost all cases is the number of signal parameters small.
Let's optimize for that by using a stack allocated buffer if we have
few parameters.
We preallocate buffers that are used after forked. That is because
malloc()/free() are not async-signal-safe and must not be used between
fork() and exec().
However, for the child process that exits without fork, valgrind wrongly
reports these buffers as leaked.
That can be suppressed with "--child-silent-after-fork=yes", but it is
cumbersome.
Work around by trying to allocate the buffers on the stack. At
least in the common cases where the pointers are small enough
so that we can reasonably do that.
If the buffers happen to be large, we still allocate them on the heap
and the problem still happens. Maybe we could have also allocated them
as thread_local, but currently glib doesn't use that.
[smcv: Cosmetic adjustments to address review comments from pwithnall]
With bash completion version lesser than 2.10, only prefix is defined
while for greater version it is datadir.
Closes#1054
Signed-off-by: Frederic Martinsons <frederic.martinsons@sigfox.com>
We format the message into a string twice, once for each byte-order,
but only return the one corresponding to the last byte-order to the
caller. This means we need to free the first one.
Signed-off-by: Simon McVittie <smcv@collabora.com>
For manual test coverage that would reproduce the bug fixed in !1902,
copy /bin/true (or any other harmless executable) to
/usr/bin/spawn-test-helper.
Signed-off-by: Simon McVittie <smcv@collabora.com>
do_exec() and g_execute() rely on being passed a NULL search path
if we intend to avoid searching the PATH, but since the refactoring
in commit 62ce66d4, this was never done. This resulted in some spawn
calls searching the PATH when it was not intended.
Spawn calls that go through the posix_spawn fast-path were unaffected.
The deprecated gtester utility, as used in GTK 3, relies on the
ability to run an executable from the current working directory by
omitting the G_SPAWN_SEARCH_PATH flag. This *mostly* worked, because
our fallback PATH ends with ".". However, if an executable of the
same name existed in /usr/bin or /bin, it would run that instead of the
intended test: in particular, GTK 3's build-time tests failed if
ImageMagick happens to be installed, because gtester would accidentally
run display(1) instead of testsuite/gdk/display.
Fixes: 62ce66d4 "gspawn: Don’t use getenv() in async-signal-safe context"
Bug-Debian: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=977961
aclocal's ability to compare the version of macros and use the latest
version relies on the serial number being incremented on every change,
or at least on every functional change.
Fixes: 6f26637e "m4macros: replace obsolete macros AC_TRY_RUN and AC_TRY_LINK in glib-2.0.m4"
Signed-off-by: Simon McVittie <smcv@collabora.com>
Running autoconf 2.70 with -Wall,error on a configure.ac that uses
AM_PATH_GLIB_2_0 gives:
configure.ac:261: warning: The macro `AC_TRY_RUN' is obsolete.
configure.ac:261: You should run autoupdate.
./lib/autoconf/general.m4:2996: AC_TRY_RUN is expanded from...
/usr/share/aclocal/glib-2.0.m4:11: AM_PATH_GLIB_2_0 is expanded from...
configure.ac:261: the top level
configure.ac:261: warning: The macro `AC_TRY_LINK' is obsolete.
configure.ac:261: You should run autoupdate.
./lib/autoconf/general.m4:2919: AC_TRY_LINK is expanded from...
/usr/share/aclocal/glib-2.0.m4:11: AM_PATH_GLIB_2_0 is expanded from...
configure.ac:261: the top level
Run autoupdate on glib-2.0.m4 to change AC_TRY_RUN and AC_TRY_LINK into
the suggested alternative, and adjust the formatting a little bit.
The macros used in the alternative existed for long enough that there
shouldn't be a problem with backwards compatibility.
Signed-off-by: Simon Marchi <simon.marchi@polymtl.ca>
I found myself wanting to know the test that is currently being run,
where e.g. __func__ would be inconvenient to use, because e.g. the place
the string was needed was not in the test case function. Using __func__
also relies on the test function itself containing the whole path, while
loosing the "/" information that is part of the test path.
- use watcher auto start flag.
- use watch_name_on_connection_with_closures.
- use an existing service name for auto start.
Closes#2011
Signed-off-by: Frederic Martinsons <frederic.martinsons@sigfox.com>
Where the early call to g_socket_set_option() fails because of
check_socket() failing due to `inited` still being FALSE.
This brings 634b692 back into working order, by fixing the regression
introduced in 39f047e.
Co-authored-by: Ole André Vadla Ravnås <oleavr@gmail.com>
