Current-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: git copies of cvs modules available

>>>>> "spz" == S P Zeidler <> writes:

   spz> We just about manage three builds a day on TNF hardware at
   spz> present.  We tend to have a bit more than three commits per
   spz> day. (On a good day, by a factor of 20).

with something like git, you could fork a shadow-repository per
developer at commit time.  Developer pushes from his local repository
to this shadow, and the build script pushes from shadow to the main
repository only after the build succeeds.  If the developer pushes
another commit after his test-build has already been scheduled onto a
CPU but before it finishes, it's his choice if he wants to abandon the
current build-test and immediately start from the beginning, or push
to a new repository that won't be granted a CPU until his current
build batch finishes.

Quite a few tools with no volunteer to write them are missing, but I
don't see that ``gate commits by compile/regress testing'' is so
fundamentally unreasonable regardless of project size.

   spz> Developers with commit access are supposed to compile-test and
   spz> test -before- commit.

so you agree it's reasonable, but think it's only reasonable if it's
done manually.  It's the same amount of work happening, though, just a
question of whether it's automatic or not, and whether it occurs in a
data center on fast bare metal with a clever batch scheduler vs inside
some ad-hoc VM on Mac OS X at 4200rpm and battery power.

Attachment: pgpUw8IFSQ9D6.pgp
Description: PGP signature

Home | Main Index | Thread Index | Old Index