On Tue, Jan 24, 2006 at 07:17:41PM +0000, Kumba wrote:
> Fix to the sigsuspend syscall so N32 userlands work without triggering
> non-fatal oopses. Patch was originally drafted by Daniel Jacobwitz.
> Problem was originally discovered via a configure test in the glib
> package. When run, the test triggered a segmentation fault and an
> oops, as well as causing the test, and ultimately, configure, to fail.
> Signed-off-by: Joshua Kinard <email@example.com>
> linux32.c | 19 -------------------
> scall64-n32.S | 4 ++--
> signal.c | 42 ++++++++++++++++++++++++++++++++++++++++++
> 3 files changed, 44 insertions(+), 21 deletions(-)
If you're going to mess around with sigsuspend, there's a better option
now: take a look at the recently added TIF_RESTORE_SIGMASK (committed
Everyone really should migrate over to that approach; it fixes (among
other things) a nasty debugging corner case and some code duplication.