[distcc] distcc and qemu-arm, happy together?

[Martin Pool]

On 6 September 2010 13:17,  <someone at boldandbusted.com> wrote:
> Howdy. I've been using distcc to spread my compile work across hosts for
> many years. However, now I'm branching into a different architecture for
> fun (ARM from x86). I'm using qemu-user, via some chroot and binfmt
> magic, to facilitate cross-compiling from an x86 box. When distcc is run
> under qemu-arm (with several distcc hosts configured), qemu always
> complains:
>
> "qemu: Unsupported syscall: 187"
>
> when distcc runs. Now, of course I understand that this is a real
> edge-case usage of distcc and that the real problem is within qemu, not
> distcc.

It's not really an edge case.  I know of other people using it. (maybe
not with qemu in particular.)

To judge from <http://www.informatik.htw-dresden.de/~beck/ASM/syscall_list.html>
that is sys_sendfile, and from memory there is either a configuration
to tell distcc not to use it, or you could easily add one.

> However, when distcc encounters this error (something to do with
> socket support), does it actually abort remote compiles and return the
> compile to localhost?

It probably runs it locally.  What does the log tell you?  It depends
a bit on whether qemu is giving ENOSYS back to distcc or something
else.

> If so, is there some way I can make distcc ignore
> these errors (it does seem to be farming compile requests out, but
> perhaps at a lower rate?) Or, is there a better way to configure distcc
> such that it doesn't use this "unsupported" method of communication when
> distributing jobs?
>
> Thanks for distcc! :) (please reply to the list, thanks)

-- 
Martin