[petsc-dev] [mpich-discuss] MPICH migration to git

[Jed Brown]

As a test for my "git-fat" extension, I liberated the large files from
PETSc's history (managing them outside the repository so that they need not
be fetched by everyone; though if you fetch them, the working tree behaves
identically to if they were in the repository). This brings the git version
of the PETSc repository down to 50MB, and the clone takes 12 seconds:

$ time git clone git at bitbucket.org:jedbrown/petsc-git-lean

Cloning into 'petsc-git-lean'...
remote: Counting objects: 297100, done.
remote: Compressing objects: 100% (67974/67974), done.
remote: Total 297100 (delta 228357), reused 297100 (delta 228357)
Receiving objects: 100% (297100/297100), 41.22 MiB | 8.71 MiB/s, done.
Resolving deltas: 100% (228357/228357), done.
12.105 real   13.472 user   2.000 sys   127.81 cpu
$ du -hs petsc-git-lean/.git
50M     petsc-git-lean/.git

The original repository (not managing anything with git-fat) is 78MB with
git. Meanwhile, the hg clone takes 10x longer and is much larger:

$ time hg clone ssh://hg@bitbucket.org/petsc/petsc-dev petsc-dev-hg

requesting all changes
adding changesets
adding manifests
adding file changes
added 25751 changesets with 99594 changes to 10045 files
updating to branch default
4296 files updated, 0 files merged, 0 files removed, 0 files unresolved
124.376 real   40.954 user   1.827 sys   34.39 cpu
$ du -hs petsc-dev-hg/.hg
178M    petsc-dev-hg/.hg



On Wed, Jan 9, 2013 at 11:02 PM, Jed Brown <jedbrown at mcs.anl.gov> wrote:

> On Wed, Jan 9, 2013 at 10:46 PM, Sean Farley <
> sean.michael.farley at gmail.com> wrote:
>
>> Jed, you have to realize that you're the only one in this thread that
>> has been disgruntled with mercurial. Even that random dude that
>> commented still doesn't like git.
>>
>> Yes, yes, git did this light-weight branching first. But, IMHO,
>> mercurial has done it in a cleaner way. And I'll take cleaner and
>> better thought out than quick and dirty any day.
>>
>
> It should be obvious that I started the thread mostly to instigate. I
> didn't expect the trolling conditions to be so good tonight. ;-D
>
> However, you'll notice quite a number of rants within our circles on G+
> (and at large) from people that used hg for a long time and haven't looked
> back since switching to git. The opposite is rare to non-existent. In the
> end, I don't think it's deeply important either way, but a lot of our
> "peer" projects have recently switched for technical reasons and it's
> potentially fewer tools to install/systems to remember. Oh, and the git
> emacs support is so much better.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20130121/37cae9e5/attachment.html>