On Wed, 3 Jan 2001, Hubert Feyrer wrote:

> On Wed, 3 Jan 2001, David Brownlee wrote:
> > 	Looks good. I'd still prefer all the patches to be put
> > 	together in one patch file - I'm more than willing to write
> > 	any tools to automatically split/rejoin/etc.
>
> IIRC I this was discussed a few times before, with no obvious outcome. I
> think we still can do that later if need arises (I don't think it really
> will).
>
	The thought would be it could reduce complexity - we could
	eliminate the patch checksums altogether as there would only
	ever be one patchfile. When there are no patches we ship a
	blank 'patches' file, in the same way as we currently have
	a blank patch-sum.

> > 	I'd be tempted to keep a single subdir for the less predictibly
> > 	named files (scripts/files). Most packages do not have them, and
> > 	can have the directory omitted.
>
> I had a look at what pkgs have which files, and most of them only have 1-2
> "misc" files, if any. I don't think it's worth keeping a subdirectory for
> them, the pkg directory itself can store them just fine most of the time.
>
	OK. Maybe a fixed prefix for all of them for consistency?
	extra-XXXX, files-XXX ?

> > 	Any milage in merging the two *-sum files?
>
> Uh, why?
> Just because we can safe a few inodes doesn't mean we should make our code
> more complex than necessary.

	They both contain md5 checksums. If we merge the generation
	targets into a single 'make checksum' would that avoid any
	extra complexity? Though see previous comment about eliminating
	patch-sum altogether.

	Note: I don't want to turn this into a protracted debate - I
	would support the reorg going ahead later this week whatever the
	state of discussion on patch and checksum files :)

		David/absolute		-- www.netbsd.org: No hype required --