Skip to contentSkip to conversation

RFC: reflog for deleted branches

DISCUSSION Sven Strickroth 5 participants 9 messages Jun 15, 2016

lore
Share & tools
lore
#0 lore raw

Hi,

an often heart question is how to restore deleted branches. Sometimes those can be restored by using the reflog of the HEAD.

But why not create a special (ref)log for deleted branches (with SHA1, date, name of the branch and maybe commit message of the last commit on the deleted branch)? This way people could just look up the SHA1 of the deleted branch and can recreate it (even if they moved the HEAD around a lot).

Signature · 4 lines
Best regards,
 Sven Strickroth
 ClamAV, a GPL anti-virus toolkit   http://www.clamav.net
 PGP key id F5A9D4C4 @ any key-server
#1 lore raw
Quoted 11 lines · Sven Strickroth

Hi,

an often heart question is how to restore deleted branches. Sometimes those can be restored by using the reflog of the HEAD.

But why not create a special (ref)log for deleted branches (with SHA1, date, name of the branch and maybe commit message of the last commit on the deleted branch)? This way people could just look up the SHA1 of the deleted branch and can recreate it (even if they moved the HEAD around a lot).

Something like this may well be needed by an SVN importer, as it's quite possible for an SVN branch to be deleted, then another branch forked from it years later. See the (brief) discussion here:

http://article.gmane.org/gmane.comp.version-control.git/195517
http://article.gmane.org/gmane.comp.version-control.git/195544

This is really Florian's area, but I expect it would be easy enough for SVN import to build on top of any reasonably general solution if you're offering :)

- Andrew

#2 lore raw
In reply to Sven Strickroth
Quoted 8 lines · Sven Strickroth

an often heart question is how to restore deleted branches. Sometimes those can be restored by using the reflog of the HEAD.

But why not create a special (ref)log for deleted branches (with SHA1, date, name of the branch and maybe commit message of the last commit on the deleted branch)? This way people could just look up the SHA1 of the deleted branch and can recreate it (even if they moved the HEAD around a lot).

The itch is legit, but I doubt a "special log for deleted branches" is a good design. Why do you limit yourself only to a single "last" commit? Wouldn't it be a better design to save away the entire reflog for the branch when a branch is deleted, possibly with expedited expire rule for such a reflog?

#3 lore raw

Am 03.05.2012 23:37 schrieb Junio C Hamano:

Quoted 2 lines

The itch is legit, but I doubt a "special log for deleted branches" is a good design. Why do you limit yourself only to a single "last" commit?

That was just the first idea ;)

Quoted 3 lines

Wouldn't it be a better design to save away the entire reflog for the branch when a branch is deleted, possibly with expedited expire rule for such a reflog?

The "normal" reflog cannot be used if you just create and then delete a branch - the reflog is empty in this case. A special "deleted" entry pointing to the SHA1 of the branch is needed. Also the question is how to handle the case that a user creates and deletes a new branch with the same name.

All in all, I (and lots of other users) would be happy if this issue could be addressed in one of the next releases. :)

Signature · 4 lines
Best regards,
 Sven Strickroth
 ClamAV, a GPL anti-virus toolkit   http://www.clamav.net
 PGP key id F5A9D4C4 @ any key-server
#4 lore raw
Quoted 2 lines · Sven Strickroth

The "normal" reflog cannot be used if you just create and then delete a branch - the reflog is empty in this case.

I somehow thought we discussed about the lack of "creation" event. Perhaps it is sufficient to create a reflog entry for such then?

#5 lore raw

Am 04.05.2012 01:05 schrieb Junio C Hamano:

Quoted 7 lines
Quoted 2 lines · Sven Strickroth

The "normal" reflog cannot be used if you just create and then delete a branch - the reflog is empty in this case.

I somehow thought we discussed about the lack of "creation" event. Perhaps it is sufficient to create a reflog entry for such then?

I suppose an entry for deletion should be sufficient.

Signature · 4 lines
Best regards,
 Sven Strickroth
 ClamAV, a GPL anti-virus toolkit   http://www.clamav.net
 PGP key id F5A9D4C4 @ any key-server
#8 lore raw
In reply to Sven Strickroth
Quoted 4 lines · Sven Strickroth

Hi,

an often heart question is how to restore deleted branches. Sometimes those can be restored by using the reflog of the HEAD.

I get by with:

    gitk --date-order $(git fsck | grep "dangling.commit" | cut -f3

-d' ') --not --branches --tags --remotes

Works for me because I usually remember *something* about the commit message so then I type in "/" and the string and in a few "Shift-downarrow"s I'm there.

#6 lore raw
In reply to Junio C Hamano

Junio C Hamano skrev 2012-05-04 01.05:

Quoted 7 lines
Quoted 2 lines · Sven Strickroth

The "normal" reflog cannot be used if you just create and then delete a branch - the reflog is empty in this case.

I somehow thought we discussed about the lack of "creation" event. Perhaps it is sufficient to create a reflog entry for such then?

We have the create event (oldd=0000...), if we just don't delete the log and add a destroy event (newid=000000).

A bit trickier with the current reflog design is the case where you delete branch foo and then create the branch foo/bar. One way would be to append a "." to the leaf in the refs directory tree. Older readers wouldn't understand that to there's a compatibility issue.

-- robin

#7 lore raw
Quoted 6 lines · Robin Rosenberg

Junio C Hamano skrev 2012-05-04 01.05:

Quoted 5 lines
Quoted 2 lines · Sven Strickroth

The "normal" reflog cannot be used if you just create and then delete a branch - the reflog is empty in this case.

Quoted 8 lines
Quoted 2 lines

I somehow thought we discussed about the lack of "creation" event. Perhaps it is sufficient to create a reflog entry for such then?

We have the create event (oldd=0000...), if we just don't delete the log and add a destroy event (newid=000000).

A bit trickier with the current reflog design is the case where you delete branch foo and then create the branch foo/bar.

That is not worth worrying about from the behavioral point of view, as end users, even on Macs and Windows, have long learned that in a hierarchical name space, you cannot have foo and foo/bar at the same time. But from the implementation point of view, that implies in this sequence:

create foo/bar, creating refs/heads/foo/bar and logs/foo/bar delete foo/bar, removing refs/heads/foo/bar but keeping logs/foo/bar create foo

the third step needs to "rm -fr logs/foo && >logs/foo". We have refs/heads/foo under lock, so presumably nobody can create a baz branch that adds logs/foo/baz to race with us while we do so, but I didn't check. Our lock taking code may need to be tweaked so that if you take lock on refs/heads/foo, nobody else can take lock on refs/heads/foo/anything, and vice versa.