[Monotone-devel] Re: monotone & CVS import

[graydon hoare]

hi,

a few people have noted recently that doing a cvs_import doesn't seem
to produce useful results. I guess I exaggerated the capabilities of
it in the manual: cvs_import is a work in progress. in monotone 0.7 it
doesn't record branch certs, so once it's imported the branches it
gives you a big ancestry tree with 0 branches. basically useless.

I've fixed this today. now when you do a cvs_import, you need to
provide a base --branch name. I recommend something along the lines of
org.domain.project. monotone will use this as a base for constructing
sub-branch names, so CVS-FUN-branch in your CVS repository would
become org.domain.project.CVS-FUN-branch in your monotone db, and
any version found which is not on a branch will just get put into
org.domain.project. this might not work *perfectly* yet, but so far
it seems to be producing somewhat sensible results on my test trees.

I also added a new variant of checkout, so you can checkout purely
by branch name, if the branch has a unique head (it'll report an
error if the branch is unmerged). so:

  monotone --db=project.db init db 
  monotone --db=project.db genkey address@hidden
  monotone --db=project.db --branch=com.foo.project cvs_import my/cvs/tree
  monotone --db=project.db --branch=com.foo.project checkout somedir

should be enough for a round trip of your CVS head. this'll show up in
0.8, but if you want it before then you'll have to build my
development version. once checked out you can of course stop issuing
the annoying --db and --branch options since they will have persisted
in the working copy, under MT/options.

if you find weird ways to make cvs_import break, please let me know.

-graydon

 (as well, I committed some speed improvements to the import cycle
  detector; unfortunately my previous post about "massive" speedups in
  large imports turned out to be slightly untrue. it's faster, but I
  don't have a measure of exactly how much faster yet. probably not as
  much as I'd hoped.)