[erlang-questions] Parallel Make in OTP

Jachym Holecek freza@REDACTED
Mon Jan 11 12:25:24 CET 2010


Hi Pete,

I completely agree with your list of properties a build system should
satisfy.

# Peter-Henry Mander 2009-12-18:
> I've heard many suggestions of replacements for Make, but a lot of
> them don't have any difference or enhancement that warrants jumping
> ship. Most seem to be changes in style more than in substance (but
> someone prove me wrong, please).
> 
> Apart from Sinan, maybe. But I've not looked at it hard enough.
> Considering the "old school" methods and process that we use in
> T-Mobile, Sinan's agile style is also a leap in culture which, I found
> from experience in the Previous Place, is much harder to push in a
> company.

I recently learnt about rebar[*]. It's a native Erlang solution and
looks pretty promising so far, although the documentation amounts to
UTSL at this point. I didn't yet have a chance to get to know it more
intimately, but that will come sooner or later. Anyway, always nice
to see your dreams come true meanwhile you were dreaming. :-)

> On a tangent: What if a declarative, rules based system was available
> in Erlang (someone's going to tell me there is one, for sure!), what
> other projects besides a drop-in replacement for a parallel,
> distributed Make could be made possible?

I've done a lot of daydreaming on the topic since we last talked and
came to the conclusion that a dependency processing engine is about
the easiest part of build system business, really. The only tricky
point is that the existence of certain arrows depends on satisfaction
status of some other arrows (yeah, that's .hrl processing). Most of
the work lies in locating resources, inferring depends and gluing it
all together...

Regards,
	-- Jachym

[*] Apologies if I'm ruining what was supposed to be undercover op:

      http://bitbucket.org/dizzyd/rebar/


More information about the erlang-questions mailing list