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

From: Robin H. Johnson <robbat2@gentoo.org>
Date: Apr 25, 2018, 20:07 UTC
Message-ID: <robbat2-20180425T173507-286694375Z@orbis-terrarum.net>
In-Reply-To: <ad4d0d66-58f4-5cab-d314-a30a50e8ad32@xiplink.com>

On Wed, Apr 25, 2018 at 11:18:26AM -0400, Marc Branchaud wrote:

Show 11 quoted lines

> > The best approach to do so is to have those people do the "touch"
> > thing in their own post-checkout hook.  People who use Git as the
> > source control system won't have to pay runtime cost of doing the
> > touch thing, and we do not have to maintain such a hook script.
> > Only those who use the "feature" would.
> 
> The post-checkout hook approach is not exactly straightforward.
> 
> Naively, it's simply
> 
> 	for F in `git diff --name-only $1 $2`; do touch "$F"; done

Even this naive attempt gets it wrong: successive files have increasing
times.  You need to capture the time at the start, and use it as the time
for the files.
  touch /tmp/ref && \
  for F in `git diff --name-only $1 $2`; do touch -r /tmp/ref "$F"; done && \
  rm /tmp/ref
(or pass a fixed time into touch).

> But consider:
> 
> * Symlinks can cause the wrong file to be touched.  (Granted, Michał's 
> proposed patch also doesn't deal with symlinks.)

Yes, it blindly touches the file, and rather than trying to do utimensat's AT_SYMLINK_NOFOLLOW flag.

Show 7 quoted lines

> Let's assume that a 
> hook can be crafted will all possible sophistication.  There are still 
> some fundamental problems:
>
> * In a "file checkout" ("git checkout -- path/to/file"), $1 and $2 are 
> identical so the above loop does nothing.  Offhand I'm not even sure how 
> a hook might get the right files in this case.

Yes, this would need to be a new hook that knows more than post-checkout presently does.

post-checkout right now runs AFTER the worktree has been updated, and only gets the refs of old/new HEAD and if the branch was changed. It does NOT know which files were actually modified, and since it

If a hook is added for that, then this behavior can be implemented in the hook. Alternatively adding a pre-checkout hook to absorb some state of the unmodified worktree (this could be a bit expensive).

Show 7 quoted lines

> * The hook has to be set up in every repo and submodule (at least until 
> something like Ævar's experiments come to fruition).
> 
> * A fresh clone can't run the hook.  This is especially important when 
> dealing with submodules.  (In one case where we were bit by this, make 
> though that half of a fresh submodule clone's files were stale, and 
> decided to re-autoconf the entire thing.)

The fresh clone case really matters for my usage, where new clones are firing in CI/CD processes.

Show 8 quoted lines

> I just don't think the hook approach can completely solve the problem.
> 
> I appreciate Ævar's concern that there are more than just two mtime 
> requests floating around.  But I think git's users are best served by a 
> built-in approach, with a config setting to control the desired mtime 
> handling (defaulting to the current behaviour).  People who want a 
> different mtime solution will at least have a clear place in the code to 
> propose a patch.

+1 as long as we can set the behavior during the clone.

-- 
Robin Hugh Johnson
Gentoo Linux: Dev, Infra Lead, Foundation Treasurer
E-Mail   : robbat2@gentoo.org
GnuPG FP : 11ACBA4F 4778E3F6 E4EDF38E B27B944E 34884E85
GnuPG FP : 7D0B3CEB E9B85B1F 825BCECF EE05E6F6 A48F6136