[Qemu-devel] Re: [PATCH v2 13/24] kvm: Set up signal mask also for !CONFIG_IOTHREAD

[Avi Kivity <avi@redhat.com>]

On 03/01/2011 10:58 AM, Jan Kiszka wrote:
> On 2011-03-01 09:39, Avi Kivity wrote:
> >  On 02/28/2011 06:54 PM, Avi Kivity wrote:
> >>  On 02/28/2011 06:49 PM, Jan Kiszka wrote:
> >>>>
> >>>>   That's what I tried, and it didn't work?!  Maybe I forgot to
> >>>  compile or
> >>>>   something.
> >>>
> >>>  Well, it maybe failed to build as qemu_kvm_init_cpu_signals became
> >>>  unused and the compiler should have bailed out? Probably it's better to
> >>>  disable it directly in the function.
> >>>
> >>
> >>  That's what I did, with #ifdefs, but brokenly (#ifndef).
> >>
> >>
> >
> >  Well it fails even with the correct #ifdef.  Maybe some later patch adds
> >  to the breakage.
>
> But when ifdef'ed out, this patch should be a nop for qemu-kvm. Indeed
> strange.

Well, there are two functions in cpus.c named 
qemu_kvm_init_cpu_signals() (an intriguing coincidence).  I #ifdefed the 
wrong one.  With the right #ifdef it works correctly.

> >
> >  This is really strange - the same test (migrate.tcp) works for Fedora
> >  and Windows XP x86.  Installation and setup of Windows XP x64 work
> >  fine.  It is only migration.tcp (when using netcat to connect to the
> >  guest) that fails.
> >
>
> Guess this has to be classically debugged. :-/ Let me know if I can help
> (though not today).

Still has to be debugged, but at least the tree is alive now.

-- 
error compiling committee.c: too many arguments to function