Ian Lynagh wrote:
On Wed, Aug 30, 2006 at 01:20:48PM +0100, Simon Marlow wrote:
Ian, what do you think?
I've had a little play with it, and it does indeed seem nice.
In particular, it makes it quite easy to extract the testsuite summary,
as well as a the bits of the log corresponding to the failures, and
making them available separately to the complete log.
I find it useful to have summaries posted to the mailing list, so we can go back
through the archives and find the point when a test started failing, that sort
of thing. I presume that's possible too?
The results of my playing are at:
http://urchin.earth.li/~ian/buildbot/
(this is just a snapshot of the real thing, which isn't accessible. Only
the links in the top left "run testsuite" box will work).
Looks cool!
I think we'll still want to have per-slave scripts in the ghc-nightly
repo, as that seems nicer than having to fiddle with the master whenever
you want to change details of one slave. Something like pre-build
(things like making mk/build.mk and setting up paths) and post-build
(putting the tarballs somewhere accessible) maybe.
Sure. I like it that our nightly builds test more than just a standard build -
e.g. including unregisterised libraries, the stage 3 build, we use the
compacting collector when building stage 3, etc. etc.
We could have a round-the-clock fast build: tune the build for speed, don't
include extralibs, just do a fast testsuite run, and perhaps do this on multiple
platforms, I think it would be a big win. I can probably find some suitable
machines here.
Cheers,
Simon
_______________________________________________
Cvs-ghc mailing list
[EMAIL PROTECTED]
http://www.haskell.org/mailman/listinfo/cvs-ghc
