Re: irc usage..

[Jakub Narebski <jnareb@gmail.com>]

Linus Torvalds wrote:
 
> [...] people _should_ realize that removing objects is very very special. 
> Whether it's done by "git prune-packed" or "git prune", that's a very 
> dangerous operations. "git prune" a lot more so than "git prune-packed", 
> of course (in fact, you should _never_ run "git prune" on a repository 
> that is active - you _will_ corrupt it)-

Would it be possible to make 'git prune' command repository corruption safe,
even if some information might be lost (like 'git add')? Or do _corruption_
mean some recoverable only information is lost? Not always one can use "one
repository per developer" workflow.


One of the solution would be to to use reader/writer lock (filesystem
semaphore), with each command modyfying repository performing locking, and
git-prune waiting on lock until noone is accessing repository. Of course
the problem is with OS and filesystems which does not support locking, and
with stale locks...

Second solution would be to [optionally] wait until no process is accessing
repository, copy repository in some safe place, [optionally] calculate
checksum, prune, [optionally] check if the repository was modified
meanwhile and either abort or repeat, and finally copy pruned repository
back.

-- 
Jakub Narebski
Warsaw, Poland