[distcc] Re: re: distributed compiler cache

[MartinPool]

On 12 Sep 2002, Joerg Beyer <j.beyer at web.de> wrote:
> MartinPool <mbp at samba.org> schrieb am 12.09.02 12:29:59:
> > On 12 Sep 2002, Joerg Beyer <j.beyer at web.de> wrote:
> > 
> > > no, not one machine each. I was not explicit about this, but my scenario 
> > > is more like this: imagine having a "reference machine" where all
> > > builds are done (a lot of software in the correct version installed,
> > > dont ask me).
> > 
> > Heh, I have the same thing at work.  It's a pain.
> 
> guess what I am talking about...

:-)

> > In that case you might as well just use 
> >   
> >   CC='ccache distcc'
> 
> but you have to wrap it into a (shell) script if you use it with
> autoconf projects.

That would be a good thing to fix up.  I don't think it requires a
rewrite. 

> > > compilation. I think there are race condiitons, right?
> > 
> > No, I don't think there's any race there.  Where do you think it is?
> 
> I am not sure, but looking at util.c:safe_open(), what is used in many
> places, e.g. stats.c: write_stats() . These calls are not locked at all,
> right? Is it save?

It looks to me like in every case where that function is called, it is
followed by a call to lock_fd().

-- 
Martin