Re: Do we need Linux/alpha and Linux/sparc versions of <bits/siginfo.h>?

[Jakub Jelinek <jakub at redhat dot com>]

> Jakub, what do you think?  Are now (with your patch and Linux
> 2.3.41pre2) all siginfo problems fixed?  Is 32bit uid support complete
> (apart from some cleaning up)?

If Linus eats my siginfo kernel patch (it clashes with your NSIGCHLD,
hopefully Linus will want to resolve that failed hunk), then hopefully
all should be ready, apart from a few missing kernel siginfo generation points.
Now the whole thing has to be heavily tested with 2.3 and 2.2 kernels.

> 
>  > IMHO, we need only separate version of sparc siginfo, plus use __WORDSIZE
>  > in the generic linux version (I think ia64, ppc64, mips64 etc. could use it
>  > as well):
> MIPS has already a special version since some constants have different
> values than the rest of the kernel :-(
>  > [...]
> New Sparc specific <bits/siginfo.h>:
>  > +
>  > +	/* SIGILL, SIGFPE, SIGSEGV, SIGBUS.  */
> You didn't add SIGEMT in the comment - is this correct?

Oopsie, you're right, should be there as well.
On the other side, the comments are pretty misleading in SIGCHLD and SIGPOLL
cases. Those structure layouts are for child deatch signal (which can be any
signal) and you can pass si_band/si_fd with signals other than SIGPOLL as
well.

Cheers,
    Jakub
___________________________________________________________________
Jakub Jelinek | jakub@redhat.com | http://sunsite.mff.cuni.cz/~jj
Linux version 2.3.40 on a sparc64 machine (1343.49 BogoMips)
___________________________________________________________________