> Aren't these statements contradictory? If you have the patch sets
> under separate subdirs, can they come from different repos? Would you
> still have to manually merge the series file? I don't totally blame
> Gabe for having the willies about this.
Yeah, it sorta gives me the willies too. Maybe what I suggest below
could be better.
> My main fear is that if all we offer as a default download is the tip,
> then if something broken gets in there and someone downloads m5 for
> the first time and it doesn't work, that's not good. Maybe Nate's
> idea of automatically advancing the stable pointer once we've run the
> fullest possible set of regressions is the right way to go, and we can
> try harder to make sure that when we see some configuration break we
> create a regression test for it.
I think this, combined with something like a testing repository which
has regular regressions run as well and is maybe similar to the
mercurial crew repository. Basically, people are allowed to push more
speculative things to the testing repository, and they'll get
regressions run on that repository. (If we do some sort of automatic
bisect, that would be awesome) If the testing repository gets totally
whacked, it's easy to blow it away and just ask people to re-add
stuff. Maybe this should be done with a mq repository instead of a
normal repository. (Is this what you had in mind steve?) This way,
when people commit stuff, they can remove it from the queue and not
figure out how to merge it. We could also alternatively come up with
some way to just have a directory of mq trees that can be individually
maintained by people, and have the regression framework automatically
do a separate regression on all of the mq trees that are there.
> That would go along with making the
> test framework a little more flexible, so we can have regressions that
> depend on files we can't export (like SPEC) that only run when
> appropriate, and maybe some staging mechanism that lets us run 1/N of
> the really long regressions every day instead of trying to do them
> all once every N days.
I am willing to work on improving the testing framework to make adding
tests easier. I'd like to base it more on SConscript stuff than just
the existence of directories. That way you could do something like
"Test('quick', mytestdir)". The result would be that it would work
with extras where we could keep the spec regressions and private
regressions. I'd love it if someone volunteered to do the bisect part
though.
Nate
_______________________________________________
m5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/m5-dev
