On 15 Apr 2008, at 20:55, Tommy Pettersson wrote:
I would probably do something like this:
- backup the modified files in bak1
- revert all unrecorded changes
- backup files modified by 4 in bak2
- obliterate 4
- unrecord 3 and amend 2 with 3's changes
- restore from bak2
- record 4 as a new patch
- restore from bak1
This method of backup-obliterate-amend-restore is sometimes
faster than multiple unrecords and interactive records,
especially if the interactive records are tedious.
Hi Tommy,
I follow the process, but how exactly would you take the backups?
Filesystem copies of the repo, followed by darcs diff? (In which
case, would step 3 be an unrecord in a backup followed by a darcs diff
to get [4] as a patch?)
I don't know if darcs makes you addicted to "a perfect history",
or if it just makes you more aware of the value of good
changesets. If we make an effort to make our code
understandable, wouldn't it make sense to also make an effort to
make our changes to it understandable?
I think probably what it does is lower the barrier to creating a good
history, and suddenly the benefits seem reasonable compared to the
cost. Not true with SVN - I used to try and keep things clean, but
more often than not I just used to dump stuff in there like a backup
device. The hunk-based rather than file-based recording, and the fact
that "darcs mv" *works* are good examples of things that makes
creating good changesets easy enough to bother.
This is my experience anyway!
Ashley
--
http://www.patchspace.co.uk/
http://aviewfromafar.net/
_______________________________________________
darcs-users mailing list
[email protected]
http://lists.osuosl.org/mailman/listinfo/darcs-users
