Machine-readable nightly output

[Edward Z. Yang]

No, I'm unfamiliar with both sets of code.  IIRC, we also have a
buildbot infrastructure, for which anyone can volunteer a machine to do
a build on, but we don't tend to do a very good job of taking care of
these machines.  Perhaps if we centralized things a bit and made it
easier for GHC devs to tweak how nightlies were being run, it would
be helpful.

Edward

Excerpts from Geoffrey Mainland's message of Thu Aug 08 21:53:41 -0400 2013:
> The nightly builds are running on machines at MSR. They're behind a
> firewall/proxy, so even getting them to send out nightly emails was a chore.
> 
> I was planning on turning off the nightlies before I leave since they
> frequently break (for example when libffi was moved into its own repo
> everything broek) and there will no longer be anyone around to fix the
> breakage. I was going to set up new nightly builds somewhere else and
> put some effort into improving them. Ideally we can move to a uniform
> build infrastructure by integrating with what is handled by
> darcs.haskell.org. I haven't taken the time to look at that code,
> though...are you familiar with it? I agree that we could use some more
> automation support.
> 
> Geoff
> 
> On 08/09/2013 01:15 AM, Edward Z. Yang wrote:
> > Hey Geoffrey,
> >
> > The nightly builds are running on your box, correct?  How
> > hard would it be to modify the nightly scripts so that
> > they also output machine-readable information about the build?
> > For example, we could then setup a fingerprint database, where
> > given some revision of head, you could find out if it built,
> > and what the hashes of all the repositories it built with were.
> >
> > Edward
> >
> > _______________________________________________
> > ghc-devs mailing list
> > ghc-devs at haskell.org
> > http://www.haskell.org/mailman/listinfo/ghc-devs
> >