From: Nathan Myers (ncm_at_[hidden])
Date: 2000-07-19 16:06:50
On Wed, Jul 19, 2000 at 08:47:05AM -0400, Beman Dawes wrote:
> At 09:34 AM 7/18/2000 -1000, Thomas Plum wrote:
> >How about adding a new shell-script component at the top-level dir
> >named e.g. structure-changes.ksh , with this intended use: All the
> >component source files are maintained with their original names, and
> >their original place in the dir structure, but after extracting a
> >complete source set, the structure-changes.ksh script is executed,
> >which re-names, removes, moves, etc, as needed ...
> That strikes me as being a useful technique, but only as a workaround for a
> version control system shortcoming. Another workarounds might include
> adding a comment to the file involved indicating its old name or old
> directory. That way these name changes will at least show up on
> diffs. But again, that's just a workaround.
It strikes me as a real problem because you can't edit the file
once it's been moved into place and then check it back into the
> It really seems a first class version control system could track
> directory and file name changes along with text changes.
CVS is universally recognized as having severe shortcomings. There
are replacements in the works. The most mature are Bitkeeper (by
Larry McVoy, bitkeeper.com) and Aegis (by Peter Miller in Australia,
www.pcug.org.au/~millerp/aegis). Both have been ported to the less
developer-friendly OSes. The trick is to persuade S/F to
run one or the other (or both) in place of CVS.
Things at S/F are chaotic, but if you were on a project with a growth
rate like theirs you would be ecstatic to have things together as much
as they do.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk