Re: [lmi] Migration to git

[Vadim Zeitlin]

On Fri, 20 May 2016 18:43:57 +0000 Greg Chicares <address@hidden> wrote:

GC> I'd rather expunge all that stuff. We're never going to go back to svn.

 This is certain, my concern was about keeping the historic references to
svn commits that make it much easier to understand the history of the code
later. We had already disagreed about this in the past, when discussing
writing commit messages, so it's probably useless to restart this
discussion again, but for me being able to do something in principle is
_qualitatively_ different from being able to do it quickly and easily.

GC> > GC> Oh...I forgot you've already done the work. Suppose, hypothetically,
GC> > GC> that I were to redo it; would our sha1sums match?
GC> > 
GC> >  No because they use the timestamp as one of the inputs and there is no 
way
GC> > you could reproduce the exact times of commit creation (which are 
different
GC> > from the author timestamps, which come from svn).
GC> 
GC> Then I won't waste my time poorly reproducing what you've already done.

 We need to reproduce it to get rid of git-svn metadata. Should I redo it
or will you?

GC> Does your migrated git repository contain lots of personal branches
GC> you've added, or is it just a copy of svn trunk? (Well, I guess I
GC> should clone it and see.)

 It does contain the branches corresponding to the open PRs and perhaps one
or two others, but you could clone only master by using --single-branch
option of git-clone. Of course, even if you do clone it with all the
branches, you can still avoid creating them in the Savannah repository by
simply not pushing them there (which won't happen by default, i.e. if you
don't use --all or --mirror options of git-push).

 But, again, it's not useful to clone it anyhow if you want to recreate it
without the metadata. The only question is whether you prefer to do it
yourself or if I should do it?

 Please let me know,
VZ