On Tue, Aug 06, 2013 at 03:58:40PM +0100, Alex Bligh wrote: > --On 6 August 2013 16:45:12 +0200 Stefan Hajnoczi > <stefa...@gmail.com> wrote: > > >>Because otherwise make check SEGVs after the patch. > > > >It wasn't clear from the patch why there would be a crash. I looked > >deeper and timerlistgroup_init() calls qemu_get_clock() indirectly, so > >we need to make sure that qemu_clocks[] is initialized to avoid a NULL > >pointer dereference. > > Actually now I recall v4 had: > > @@ -215,6 +216,12 @@ AioContext *aio_context_new(void) > aio_set_event_notifier(ctx, &ctx->notifier, > (EventNotifierHandler *) > event_notifier_test_and_clear, NULL); > + /* Assert if we don't have rt_clock yet. If you see this assertion > + * it means you are using AioContext without having first called > + * init_clocks() in main(). > + */ > + assert(rt_clock); > + ctx->tl = qemu_new_timerlist(rt_clock); > > The equivalent in v7 would be an assert in timerlist_new_from_clock > to check 'clock' is non-NULL. I shall put that in as the reason for > this SEGV is non-obvious.
Nice, the comment makes the SEGV clear.
