Michael Schwendt wrote:
On Thu, 29 Nov 2007 13:47:16 +0100, Petr Machata wrote:
> Agreed. By then the upstream could disappear, migrate to whatever hip
> vc system there is in five years, etc. And maybe I fail to see the
> intent behind using the date in release tag of vcs-checkouted package.
> Perhaps it's just an arbitrary identifier. But if the intent is
> anything close to making a life easier for whoever reconstructs the
> srpm, using the release number makes more sense than the date alone.
The instructions on how to check out the source from cvs/svn/whatever
should be put into comments in the spec file. Don't attempt at squeezing
svn revisions (and silimar identifiers) into the package file name.
Again, that depends on what's the intent behind placing the date into
release tag. From Toshio's mail I know what's the rationale behind
using the date, so I can agree with you at this point.
Originally, the date in the file name has been used as (1)
about the age of a packaged post/pre-release _snapshot_ and (2) as the
least-significant portion of %release (the newer date of a newer checkout
would make the built package an upgrade without the need to bump %release
Currently, you have to bump "base" release tag in addition to changing
the date portion.
If todays policies say you must add a date and SCM id like
they ought to be revisited. ;)
Yes, snapshot packages should, under the naming guideline, contain the
date and the string "cvs". And, as has been pointed out to me, it's
even allowed to append tree release or (part of) git tree ID to that string.