On Sun, 20 Mar 1994, Charlie Brady wrote:
> On Thu, 17 Mar 1994, Bill Mitchell wrote:
> I think that Bill's reasoning is sound (apart from the "vi" to "emacs"
> change). The only problem that I have with it is that it might be
> ambiguous, whereas a context diff is definitive.. Perhaps both
Definitive, but not comprehensive. In the case of less-177, all the
changes show up in files which are created by running the configure
script, and by doing further linux-specific edits to files created
by the script. The context diff reports the existance of these new
files, but not their contents, and not what linux-specific configuration
script responses and subsequent linux-specific file edits produced them.
> need to exist, the context diff so that so that anyone can easily rebuild
> exactly the distributed binary, and the CHANGES file as a resource to a
> subsequent developer.
The context diff doesn't contain sufficient information to reproduce the
distributed binary starting with a standard source distribution, at least
not in the case of less-177. And less-177 is a pretty simple case. The
CHANGES file would provide this missing information. (CHANGES.debian.version
might have been a bad choice for the file name. Perhaps README instead of
CHANGES would be better).
> In any case, wherever the CHANGES file says, "edit file so-and-so", a
> context diff is appropriate.
But if file so-and-so was created by a configuration script and did not
exist in the original source distribution, the context diff only reports
the existance of this new file -- not its contents and not what
configuration script responses and edits went into producing it.
> Is Bill suggesting that the CHANGES file is distributed with the binary
> distribution? If so, why and where will it be installed?
No. It's only of interest if the binaries are being rebuilt from sources.
It's intended as an aid when a debian package distribution which had been
maintained by one maintainer is taken over by another maintainer, or for
anyone else interested in building his own binaries starting with debian
Ian't guidelines say that debian package sources should (strive to) build
and install completely with a simple "make install", so they'd contain
the complete files, and the context diffs not reporting their contents
wouldn't be a problem. I'm thinking that information on how these
files were produced, and perhaps some remarks on the reasoning behind the
configuration choices which led to them, might be useful.
email@example.com (Bill Mitchell)