[m-rev.] Converting to Git.

[Julien Fischer]

On Wed, Jan 2, 2013 at 10:25 AM, Paul Bone <paul at bone.id.au> wrote:
> On Mon, Dec 31, 2012 at 03:44:50PM +1100, Peter Wang wrote:
>> On Mon, 31 Dec 2012 14:57:21 +1100, Paul Bone <paul at bone.id.au> wrote:
>> > On Mon, Dec 31, 2012 at 02:07:20PM +1100, Julien Fischer wrote:
>> > > Hi Paul,
>> > >
>> > > On Sat, Dec 29, 2012 at 6:24 PM, Paul Bone <paul at bone.id.au> wrote:
>> > > >
>> > > > Mercury is now converted to git and uploaded to github.
>> > >
>> > > I notice you've shifted tests directory into the main repository
>> > > (which is fine),
>> > > however we still need the benchmarks directory from the old CVS repository
>> > > as well.  (IMO, it may as well go in the main repository as well.)
>> >
>> > I can re-do the conversion so that we get the history of benchmarks inline
>> > with the main repository (the way I've done for tests).  Or I can just add
>> > it at this point, throwing away the benchmark's repos history.
>> >
>> > The former is more work, especially if people have already started using
>> > this.
>>
>> If you will be redoing the conversion, can you improve it a bit:
>>
>> - add author names and email addresses
>> - some summary lines unnecessarily duplicate the first line, e.g. 5eb9c3
>> - some commits have a filename as the summary, e.g. 976dae
>> - some commits have two blank lines following the summary
>
> Unless there is a really compelling reason to re-do the conversion (if the
> history of benchmarks/ was important for example) then I'd re-do it.
> Otherwise I think the current conversion "is good enough".

I think so too, although it seems to have resurrected a number of files that
had been deleted in CVS, e.g.

    compiler/callgraph.m
    compiler/implication.m
    compiler/ml_dense_switch.m
    compiler/vn_livemap.m

all shouldn't be there.

> These changes would be good, but at least to us (or anyone looking at the
> authors page) we know who 'pbone" "zs" "wangp" etc are, and the formatting
> doesn't matter much as all the right information is there.
>
>> Otherwise, I'll fix those up next week.
>
> That would create a history with new hash IDs, and therefore will break any
> checkouts that others have made.  I know a few people who have already
> cloned the repository, despite us not saying it's ready, so I'd rather not
> do this.
>
> How do others feel about using the current (good enough) conversion?
> Or fixing up the issues identified by Peter?

I don't care either way -- I do have changes that I would like to commit at
some point, so if it's ready to go now then that would be good.

Cheers,
Julien.