On Mon, 2004-07-05 at 19:15, Wolfgang Grandegger wrote: > On 07/05/2004 06:46 PM Philippe Gerum wrote: > > On Mon, 2004-07-05 at 17:52, Wolfgang Grandegger wrote: > >> On 07/05/2004 05:22 PM Philippe Gerum wrote: > >> > On Mon, 2004-07-05 at 17:17, Gilles Chanteperdrix wrote: > >> >> Wolfgang Grandegger wrote: > >> >> > $ grep LATENCY config.log > >> >> > #define CONFIG_RTAI_SCHED_8254_LATENCY > >> >> > #define CONFIG_RTAI_SCHED_APIC_LATENCY > >> >> > >> >> These constants do not have the same name in Kconfig and configure.in, > >> >> hence the problems you see. > >> >> > >> >> I saw and fixed the same problem for CONFIG_RTAI_HW_NRCPUS but did not > >> >> see that some other constants had the problem. > >> > > >> > RTAI_HW_* is fusion stuff. I checked in vesuvio, and all seems to be ok > >> > in this respect. However, PPC did not define those symbols in Kconfig; > >> > Wolfgang, did you add them as strings (and not bools) there? > >> > >> I copied and modified the definition: > >> > >> config RTAI_SCHED_8254_LATENCY > >> string "8254 tuning latency (ns)" > >> default 4700 > >> > >> from i386/Kconfig. But the problem is that in "configure.in" around the > >> latency value checks there is: > >> > >> if test "$CONFIG_X86" = y; then > >> ... > >> fi > >> > > > > Ok, I understand now. I overlooked that. > > > >> Well, at least CONFIG_RTAI_SCHED_8254_LATENCY is useful as general timer > >> latency value for non-x86 archs as well because they are used indirectly > >> by the scheduler. > >> > > > > Ok, so let's kill the X86 test then. I've removed it in the CVS. > > D'accord!
Parfait ! :o) > Because even with the x86 test the CONFIG_RTAI_SCHED_*_LATENCY > definitions show up in rtai_config.h. Because autoconf has to produce the rtai_config.h template (i.e. through the bootstrap script) before any configuration takes place, so all symbols must be dumped. We just don't know which one will be useful when configuring at this point. > Now I'm puzzled with another > configuration problem. After a "make distclean" I always get "N" as > default for RTAI_SHM: > > $ make ARCH=ppc CROSS_COMPILE=ppc_8xx- config > ... > Shared memory (RTAI_SHM) [N/m/y/?] > > As for i386, the corresponding Kconfig entry is: > > config RTAI_SHM > tristate "Shared memory" > default m > help > > Any idea why? > The defaults are taken from arch/*/defconfig. The one which is currently in effect for PPC disables SHM. Copying .rtai_config to defconfig after having defined the default profile you think is the best for PPC would do it. > Another minor issue: RTAI_MATH should depend on RTAI_FPU_SUPPORT, IMHO. > Ack. > Thanks. > > Wolfgang. > > _______________________________________________ > Rtai-dev mailing list > [EMAIL PROTECTED] > https://mail.gna.org/listinfo/rtai-dev -- Philippe.
