[distcc] Re: Announcing distccKNOPPIX

Alexandre Oliva oliva at lsd.ic.unicamp.br
Tue Jul 8 04:27:36 GMT 2003


On Jul  8, 2003, Martin Pool <mbp at samba.org> wrote:

> Yes, but it remembers that the host is down, and won't try it for
> another minute.

Yeah, this was a great improvement, thanks for that!

> I'd like to see somebody right a standalone program that does mDNS or
> an nmap scan and creates a ~/.distcc/hosts file based on the results.

~/.distcc/hosts doesn't quite cut it for me.  Because of the way I use
cross, native or cross-native (*) toolchains, the same set of hosts
may or may not work, depending on what CC I'm using, for example, the
system's GCC, that varies depending on the exact version of Red Hat
Linux, sometimes betas or even rawhide I'm running on each of my
hosts; or GNU Pro natives or crosses that work identically on all of
them.  It would be nice if I could express this kind of dependency on
in DISTCC_HOSTS or ~/.distcc/hosts somehow.

That said, if instead of rebuilding ~/.distcc/hosts, it would just
mark in a separate file certain hosts as unusable for the time being,
or unusable for some particular CC, or something like that, I could
probably cobble up a shell script or something to that effect.

The need for per-host validation of CC compatibility reminded me of an
issue I had when trying to play with distcc over (fsh/)ssh (fsh would
cut down connection set up time) was that, since it wouldn't set up
the PATH like a login session would, it would end up not finding
soft-links to compilers that I had in my PATH on the build machine,
but that would be correctly found in the distccd environment when I
started distccd in a login session.  I couldn't find a clean solution
for this, so I ended up giving up on distcc over ssh :-(  Anyway, if
distcc (or distccd?) could source a script to set up an environment
before starting the compiler on the remote host, this would help.
~/.distcc/env?  Alternatively, it could support mapping compiler
pathnames to host-dependent full pathnames, but this doesn't sound as
useful or generic as a server-side env setting.

Thoughts?

-- 
Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer                 aoliva@{redhat.com, gcc.gnu.org}
CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist                Professional serial bug killer



More information about the distcc mailing list