[distcc] distcc with xinetd... how?

[Martin Pool]

On 13 Aug 2002, Brian Hall <brianwhall at pcisys.net> wrote:

> So, looks like ~150% speedup, with 25% less local CPU loading. Nice.

THanks for that.

> BTW, I notice some parallel distcc compiles seem to fail due to race
> conditions- i.e. a compile is handed off to a slower processor, but the
> faster local processor catches up and tries to write the same output
> file. This condition is probably more obvious on my compile cluster due
> to the wide range of CPU speeds.

My initial suspicion would be that this is actually a concurrency bug
in the makefile, not in distcc.  Many Makefiles seem to have latent
bugs because they've never been built with so many CPUs.

Could you perhaps try to reproduce it with DISTCC_LOG and
DISTCC_VERBOSE set to capture the output and let me see what's
happening in there?

Thanks,
-- 
Martin 

<shawnm> why is the airport in Toronto called yyz?
<tpot_> xyzzy was too long