Mats Lidell writes:
>>>>> Raymond Toy <rtoy(a)google.com> writes:
> What platforms would those be?
I don't know if we have an official list of platforms we have promised
to support. My guess at it would be linux in different flavours,
Windows, and OSX. If applicable 32- and 64-bit versions of those
above.
We definitely support (Open)Solaris and the *BSDs; there may be some
others. Of the usual suspects, I think maybe Mike S runs FreeBSD, R
Sparapani runs Solaris. I might be able to supply a NetBSD build
slave.
More important than wordsize to Mac OS X is system version (personally
I'd at least like to see Snow Leopard and Lion, Lion is supposed to
cause trouble for GNU Emacs and I run anything-but-Lion versions going
back to Tiger, but I know Snow Leopard is in fairly common use).
However, I think that having lots and lots of buildbot slaves is not
necessarily a good thing. What I see on the Python buildbot farm is
that keeping the bots happy takes a lot of effort. I don't think
that's a profitable place for us to put in much effort. Rather I
think we should focus on common configurations that exercise large
blocks of code that differ and aren't really optional. Ie, for the
moment I see the purpose of having the bots is an early warning
system of breakage so we can say to people, "Sorry about the trouble,
but that's probably been fixed already. Try 'hg pull -u', rebuild,
and let us know if you still have a problem."
So, for example, although I've heard that people do have trouble on
Lion, it's not obvious to me that we need more than a Linux build.
Most typos that break the build on Mac OS X will break most POSIX-y
systems (possibly including Cygwin). OTOH, we do need a bot for
(native) Windows, because that code base has a huge component that's
nothing like the Unix counterpart, and APIs that change will break the
build. I would say that on POSIX-y systems we should probably have
both Xt/Athena-ish and GTK+ bots, though, because even though much of
the low-level graphical code is very similar, both the event loops and
the widget management code differ substantially and would be subject
to API skew issues.
If we get "other" bots, maybe we should have a two-tier system. There
would be a set of core bots (as in the previous paragraph) where we
run both a build and the tests (but only report regressions by
default), and the "other" class of bots would be configured to just
report build breakage.
_______________________________________________
XEmacs-Beta mailing list
XEmacs-Beta(a)xemacs.org
http://lists.xemacs.org/mailman/listinfo/xemacs-beta