This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Repository for the conversion machinery


Jason Merrill <jason@redhat.com>:
> > I've used both git-svn (sometimes with git filter-branch) and reposurgeon
> > for repository conversions.  My experience is that if there's anything at
> > all complicated or messy about the history, using git-svn for the
> > conversion is not a good idea, so I don't think that's an attractive
> > option at all.
> 
> But we're already using git-svn, and it's fine.

My hair stands on end when I hear anyone say that.  git-svn is tolerable,
if a bit flaky, for live gatewaying. It is *dangerous* for whole-history
conversions.  I wrote a public-service announcement about this a while back.

Don’t do svn-to-git repository conversions with git-svn!
      http://esr.ibiblio.org/?p=6778

It should be more widely known that git-svn is bad in this mode; it's
not because on small, linear or near-linear repos without operator
errors or CVS-conversion scar tissues it happens to work reasonably
well. GCC is not in this category.

> Specifically, reposurgeon doesn't like subdirectory branches much more
> than git-svn does, though I was able to work around that with
> branchify (and some fixes in reposurgeon).  It discards branches with
> no additional commits, and keeps other branches that were deleted in
> SVN, though I was able to work around this with a postprocessing
> script.
> 
> Most significantly, it fails to handle some commits (I noticed r137307
> and r131989), leading to incorrect file content at the top of the
> affected branches.  This is the issue that I didn't hear back from
> Eric about, which led me to consider other approaches.  Fortunately it
> marked them (as well as many others) with emptycommit tags for review.
> 
> But if you want to do it with reposurgeon, I won't complain.  I've
> pushed my WIP to https://github.com/jicama/gcc-reposurgeon

This is my fault. I think you hit that bug right around one of the times I
was being most distracted by NTPsec, and I somehow formed the mistaken
impression you had gotten past it.

I'll get back on this.  I still have your bug report in my back mail; the
first thing I need to do is fix that.  Can't promise immediate action, I'm
preparing for the 1.0 release of NTPsec (spent the last week mostly
clearing our issue list) but now that I know this is unfinished I won't
drop it until it is.
-- 
		<a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]