Crash interception/debugging systems like Apport or ABRT capture core dumps for
later crash analysis. However, if a program exits with an assertion failure,
the core dump is not useful since the assertion message is only printed to
stderr.
glibc recently got a patch which stores the message of assert() into the
__abort_msg global variable.
(http://sourceware.org/git/?p=glibc.git;a=commitdiff;h=48dcd0ba)
That works fine for programs which actually use the standard C assert() macro.
This patch adds the same functionality for glib's assertion tests. If we are
building against a glibc which already has __abort_msg (2.11 and later, or
backported above git commit), use that, otherwise put it into our own field
__glib_assert_msg.
Usage:
$ cat test.c
#include <glib.h>
int main() {
g_assert(1 < 0);
return 0;
}
$ ./test
**ERROR:test.c:5:main: assertion failed: (1 < 0)
Aborted (Core dumped)
$ gdb --batch --ex 'print (char*) __abort_msg' ./test core
[...]
$1 = 0x93bf028 "ERROR:test.c:5:main: assertion failed: (1 < 0)"
https://bugzilla.gnome.org/show_bug.cgi?id=594872
Not only is the default implementation broken (it causes infinite recursion
as seen in bug #603982), but its also worthless. If we just fall back on the
default stream operations we automatically get async version based on
the sync filter stream operations, which is what we want.
Note: Since we export types with Iface in the name rather than
Interface we have to use some typedefs to make this work. New
interfaces should probably use Interface as the public name.
This works around the need to take a custom mutex twice and add the
object to a GSList of objects that are currently in construction for the
common case. Only when the constructor is overwritten do we use the
previous behavior and allow things like singleton objects.
The only slightly incompatible change is that previously, it was ok to
call g_object_set() on construct-only properties while the object was
initialized. This will now fail. If that behavior is needed, setting a
custom constructor that just chains up will reenable this functionality.
https://bugzilla.gnome.org/show_bug.cgi?id=557151
g_type_default_interface_peek() and g_type_value_table_peek() don't need
to acquire read locks anymore when they test the refcount instead of
node->data.
