Re: [Qemu-devel] [PATCH 05/10 v4] bsd-user: Implement new syscall print_sysarch and add strace support

[Sean Bruno <sbruno@ignoranthack.me>]

On Sun, 2014-06-15 at 20:20 +0100, Peter Maydell wrote:
> On 15 June 2014 16:33, Sean Bruno <sbruno@ignoranthack.me> wrote:
> > I think there's some confusion between the strace support for the new
> > print_sysarch() and the existing syscall do_freebsd_sysarch().
> 
> Only because you've put parts of both in the same patch :-)
> 

Oh, I didn't say *where* the confusion was.  It is most definitely on my
side.  :-)

> > If I follow the code, the existing do_freebsd_sysarch() syscall is a
> > programtical way of figuring out what arch is running.  Whereas
> > print_sysarch() spams the arch into your strace output.
> 
> The existing do_freebsd_sysarch() is a bunch of functions in
> syscall.c, with TARGET_* ifdefs selecting which one you get.
> This patch seems to be attempting to change that to having
> the per-arch implementations in the per-arch files. That's a
> good idea, but this patch is only doing half of the job -- you
> need to remove the old implementations and wire up the new.
> Really the changes to the implementation and to the strace
> support should go in separate patches.
> 
> > Bearing that in mind, I think that the changes here are indeed correct
> > for this patchset.
> 
> I still disagree here. Look at the TARGET_I386 implementation
> of do_freebsd_sysarch() in the existing syscall.c, and at the
> new function do_freebsd_arch_sysarch() you've added in this patch in
> bsd-user/x86_64/target_arch_sysarch.h. They're basically
> identical -- this should be a code-move change, but you've
> only got the 'add new version', not the 'and remove the old'.
> 
> I think if you remove all the do_freebsd_arch_sysarch()
> functions from this patch you're left with just the strace
> support (the strace related functions all have 'print' in
> their names).
> 
> thanks
> -- PMM


Ok, more staring required.

sean