On Thu, May 24, 2018 at 10:52:26AM -0500, Eric Blake wrote: > On 05/24/2018 10:38 AM, Michael S. Tsirkin wrote: > > Add more checks on how did QEMU exit. > > > > Legal ways to exit right now: > > - exit(0) or return from main > > - kill(SIGTERM) - sent by testing infrastructure > > > > Signed-off-by: Michael S. Tsirkin <[email protected]> > > --- > > > > Changes from v1: > > - drop SIGTERM as suggested by Eric > > > > > +++ b/tests/libqtest.c > > @@ -110,7 +110,12 @@ static void kill_qemu(QTestState *s) > > pid = waitpid(s->qemu_pid, &wstatus, 0); > > if (pid == s->qemu_pid && WIFSIGNALED(wstatus)) { > > Wait a moment. If WIFSIGNALED() is true... > > > + /* Core dump is never OK */ > > assert(!WCOREDUMP(wstatus)); > > + /* Must exit normally */ > > + assert(WIFEXITED(wstatus)); > > ...then WIFEXITED() is false. This is bogus. > > > + /* If exited normally - check exit status */ > > + assert(!WIFEXITED(wstatus) || !WEXITSTATUS(wstatus)); > > And you have some redundancy - !WIFEXITED() is not possible if you just > asserted WIFEXITED(). > > Better would be: > > if (pid == s->qemu_pid) { > /* > * Since sending SIGTERM turns into a normal exit, we want to flag > * any non-normal exit, whether or not it dumped core, as a test > * failure (even if it was a SIGKILL from someone desperate to stop > * the testsuite). > */ > assert(WIFEXITED(wstatus) && !WEXITSTATUS(wstatus)); > } > > Also, since waitpid() can only return either s->qemu_pid or -1 as we aren't > using WNOHANG,
> it may also be worth asserting that if pid == -1, we either > have EAGAIN (but why aren't we looping in that case?) I don't think waitpid can return EAGAIN > or ECHILD. Right but checking for known failures explicitly is helpful for debugging. > -- > Eric Blake, Principal Software Engineer > Red Hat, Inc. +1-919-301-3266 > Virtualization: qemu.org | libvirt.org
