Re: empty TIMESTAMP, stamp-vcl

[Gary V. Vaughan]

Hallo Bruno,

On 28 Mar 2007, at 22:29, Bruno Haible wrote:
> Eric Blake wrote:
> > Now that you mention it, it would be nice if every invocation of
> > 'gnulib-tool --update' or 'gnulib-tool --import' would touch a  
> > timestamp
> > file that could then be incorporated into the project.
>
> How would this help? It's not the date of the latest gnulib-tool  
> invocation
> that matters. Nor is it the date of the latest modified file in the  
> gnulib
> checkout. But rather the collection of the revisions of all checked- 
> out files.
> I.e. the 'git' tree serial number.

Or any other means of being able to get back to that exact collection of
revisions.  For CVS m4 and libtool, which requires a ChangeLog entry for
every checkin, the CVS revision number of the ChangeLog file serves that
purpose adequately.

> Therefore would it help if gnulib-tool copies the contents of
> $gnulibdir/.git/refs/heads/master into the generated gnulib-cache.m4?

I am not yet familiar with git.  This will only work if the gnulib tree
was checked out with git, right?  Assuming that's all that's needed to
tcheck out the identical gnulib files again at some future point, it
certainly solves half the problem.  Is there an equivalent way to  
provide
that information when CVS was used to get the gnulib repository used at
m4 build time? (If the answer is: "no, you'll have to use git if you  
need
to track releases like that", I'm willing to write an Autoconf macro  
that
mandatesa git gnulib checkout in order to build m4).

Cheers,
        Gary
--
  ())_.              Email me: address@hidden
  ( '/           Read my blog: http://blog.azazil.net
  / )=         ...and my book: http://sources.redhat.com/autobook
`(_~)_ Join my AGLOCO Network: http://www.agloco.com/r/BBBS7912

PGP.sig
Description: This is a digitally signed message part