2009/5/28 Marijn Schouten (hkBst) <hk...@gentoo.org>: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Piotr Jaroszyński wrote: >> 2009/5/28 Ulrich Mueller <u...@gentoo.org>: >>>>>>>> On Thu, 28 May 2009, Tiziano Müller wrote: >>>>> ${PORTDIR}/app-misc/foo/foo-1a_live.ebuild >>>>> ${PORTDIR}/app-misc/foo-1a/foo-1a-live.ebuild >>>> you probably mean: >>>> ${PORTDIR}/app-misc/foo-1a/foo-1a.live.ebuild >>> No, I mean what I had written, namely to use an underscore as >>> separator, i.e., "_live". But when the version is just "live" alone, >>> one would suppress the underscore for aesthetic reasons, i.e. instead >>> of "foo-1a-_live" it would be "foo-1a-live". >>> >>>> but how would their vdb or binpkg names be unique? >>>> vdb for example: >>>> app-misc/foo-1a_live for app-misc/foo >>> PN=foo, PV=1a_live => app-misc/foo-1a_live >>> >>>> app-misc/foo-1a_live for app-misc/foo-1a >>> PN=foo-1a, PV=live => app-misc/foo-1a-live >>> >>>> am I missing something? >>> Everything is easy, if you keep the following rule in mind: >>> >>>>> With our current versioning scheme the rule is very simple: ${P} is >>>>> split into ${PN} and ${PV} at the last hyphen. This can be done in >>>>> a straight forward way by regexp matching, and I would really hate >>>>> to lose this nice property. >>>> I don't understand why this property is important. Can you please >>>> explain? >>> See above, it automatically avoids any ambiguities in splitting P into >>> PN and PV. And look at function "pkgsplit" in Portage: It can just >>> treat PV as an opaque string. >>> >>> What would be the advantage to use a hyphen instead of an underscore? >> >> Mainly the thing you observed yourself - foo_live is a bit >> inconsistent with current versions. > > Ulrich is proposing foo-live if live is the entire version, foo_live is not a > legal `package name and version'. (It could be a package name though.)
I know, it's also inconsistent. Anyway that's really not that important. We could forbid package names that end with a -$PV where $PV is a valid version spec. That would make package names like foo-1b invalid (didn't see anything named like that in the tree anyway). >> The case you mention can be avoided with another restriction in PMS. >> Buut we might as well go all the way and change the version separator >> to -- or something, which would be the most flexible. > > That would also be a good solution though we don't seem to need it yet. It > would > also entail compatibility issues. Not with g55. -- Best Regards, Piotr Jaroszyński