Re: [Gnu-arch-users] Good News about GNU Arch!

[Alfred M\. Szmidt]

   I see what you're saying -- I though you were just suggesting to
   throw away the top-level of the filenames used in the diff header.

Nope, though I wouldn't mind throwing it away since I find it
cluttered. ;-)

   What you're _really_ suggesting could be useful.

I was a bit unclear yes, sorry for that.  So that others also
understand, what I was suggesting was adding `this patch is from this
changeset' blurb right after the timestamp.  That is the following:

--- FILE DATE PREV-CHANGESET
+++ FILE DATE CHANGESET

Where FILE and DATE are what they are currently.  CHANGESET would be
empty if one just does `tla changes --diffs' on a already modifed tree
with changes that haven't been commited.  CHANGESET and PREV-CHANGESET
don't need to be just the patch-level, they could be a full archive
name too, or just a partial.  I think that I would prefer a full one
instead of just seeing `patch-3'.  But some people might complain
about the length (just as they complain about `funny chars' and `long
filenames').

      --- tla--devo--1.3--patch-3/main.c      13 Jun 2005 00:22:11 -0000
      +++ tla--devo--1.3--patch-4/main.c      28 Oct 2005 09:04:31 -0000

I'd prefer to have that as "real" meta data in the patch header
instead of having it in embedded in the filename.  But each to their
own.

   or thereabouts.  I suppose there might be scripts depending on the
   presence of "mod/" or "orig/" but those I've written seem to
   generally just do patch/whatever -p1, so only care that there is a
   single meta-level of some sort.

Not knowing tla to well, must it have a single meta-level in the patch
header?