[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Gnu-arch-users] What is the use of {arch}/,,inode-sigs
From: |
Miles Bader |
Subject: |
Re: [Gnu-arch-users] What is the use of {arch}/,,inode-sigs |
Date: |
Thu, 11 Mar 2004 19:15:39 -0500 |
User-agent: |
Mutt/1.3.28i |
On Tue, Mar 09, 2004 at 05:37:33PM -0500, Stefan Monnier wrote:
> I thought they were similar to the CVS/Entries file: used to figure out
> which files have been changed without doing actual file-compares with the
> pristine copies (or at least avoid a large number of such compares).
Yes, that is what they're for.
> But the files there seem to all be "out of date" (e.g. there's data for rev
> 9 and 13, but I'm currently at rev 14).
Not every tla command updates them; I think currently only `commit' and `get'
do (and `update' and `replay' don't).
This situation is not strictly speaking a bug -- inode sigs are not required,
just an optimization -- but it would in fact be very nice for more commands
to update them.
I think the current behavior is due to the fact doing it for update/replay is
not entirely straightforward: whereas `commit' can just snapshot the tree
state and be certain that the snapshot corresponds to the just committed
patch-XX, replay/update may be applied to a tree containing uncommitted
changes; it might be possible (though perhaps non-trivial) to incrementally
modify an existing inode-sig in line with patches applied, but even then you
can't really be sure -- what if a patched file contains local changes?
Patches accepted, I'm sure. :-]
-miles
--
"Most attacks seem to take place at night, during a rainstorm, uphill,
where four map sheets join." -- Anon. British Officer in WW I