Re: [RFC PATCH] checkout: Force matching mtime between files

[Junio C Hamano <gitster@pobox.com>]

Elijah Newren <newren@gmail.com> writes:

> Here's a crazy idea -- maybe instead of a list of pathspecs you just
> provide the timestamp of when git checkout started.  Then the hook
> could walk the tree, find all files with modification times at least
> that late, and modify them all back to the the timestamp of when the
> git checkout started.
>
> Would that be enough?  Is that too crazy?
>
> Sure, people could concurrently edit a file or run another program
> that modified files, but if you're doing that you're already playing
> race games with whether your next incremental build is going to be
> able to be correct.  (Some (annoying) IDEs explicitly lock you out
> from editing files during a build to attempt to avoid this very
> problem.)
>
> That does leave one other caveat: If people intentionally do really
> weird stuff with having files with modification timestamps far in the
> future.  However, it seems likely that the group of people doing that,
> if non-zero in number, is likely to be dis-joint with the group of
> folks that want this special
> uniform-timestamp-across-files-in-a-checkout behavior.

These two groups may share the same degree of insanity ;-)

But the single timestamp idea certainly sounds workable, except that
care must be taken to make sure we really grab the fs timestamp (it
is not uncommon for ">F; stat F" to yield quite different time from
"date", when the filesystem is on a remote box).