[Savannah-hackers-public] Re: log_accum at Savannah

[Sylvain Beucler]

On Mon, May 08, 2006 at 10:55:19AM -0400, Derek R. Price wrote:
> Sylvain Beucler wrote:
> > I am eventually working on merging our versions of log_accum and
> > commit_prep with the ones at
> > http://cvs.sv.gnu.org/viewcvs/ccvs/contrib/?root=cvs
> > 
> > After some work to support our features, it occurs to me that the
> > contrib/ scripts as-are cannot work. At least,
> 
> I think it should be able to.  It's possible that I ported a recent
> feature or two without testing thoroughly, but it worked not very long
> ago and I will be happy to help solve any remaining problems.
> 
> > - the code for the new CVS loginfo options pattern (%{sVv} instead of
> >   %1{sVv}) does not support getting the current directory as the first
> >   non-option argument - so log_accum never think it worked on the last
> >   directory given by commit_prep, and never send mail.
> 
> This is not true.  The current directory is now available to all trigger
> hooks via via the "%p" format string (reference:
> <http://ximbiot.com/cvs/manual/cvs-1.12.13/cvs_18.html#SEC185>).  The
> documentation simply moved to the common page about the trigger scripts
> since %p is now one of the standard format strings interpreted by all
> the trigger hooks.

Sorry, I mean: log_accum, when $UseNewInfoFmtStrings = 1, assumed that
only %{sVv} was passed, and thus could not know what the current
directory was. I modified it to assume that %p %{sVv} is passed.

> > I've pretty much fixed those, but I wonder whether there is a point on
> > working in your versions of log_accum/commit_prep if they are
> > broken. Or maybe I totally missed something :)
> 
> I don't think it is broken, at least not badly.  If you wouldn't mind
> sending me the patch you've already come up with, I'll be happy to apply
> it and work out any other bugs I can find on this end before bouncing it
> back to you.

I attach a diff of my current versions. I'm still working on them though.

-- 
Sylvain