On 02/13/2014 07:08 PM, Philippe Gerum wrote: > On 02/13/2014 06:51 PM, Gilles Chanteperdrix wrote: >> On 02/13/2014 06:48 PM, Philippe Gerum wrote: >>> We'd need the patch below at the very least to fix the obvious sched >>> slot init issue, but there is more actually. rthal_timer_request() makes >>> the same broken assumption that CPU0 will be part of the rt set for all >>> SMP archs the same way. >> >> I am the one who made the broken assumption... I simply assumed all cpus >> were real-time at that time. And made the same assumption when writing >> the switchtest test. >> > > Nope, the rthal_timer_request() is mine, I recognize bugs I committed > when I see them: they most often come back and haunt me like a curse, > long after I forgot the reason for writing the broken code. But hey, > I'll happily leave you all the rest! > I remember faintly changing a test counting the number of timers requested with a test with cpu == 0. But looking in ksrc/arch/*/hal.c, I see that the test for cpu_timers_requested is still there. Must have dreamed.
-- Gilles. _______________________________________________ Xenomai mailing list Xenomai@xenomai.org http://www.xenomai.org/mailman/listinfo/xenomai