On Mon, Jun 11, 2012 at 12:34:16PM +0100, Matthew Seaman wrote: > On 11/06/2012 11:32, Baptiste Daroussin wrote: > >> UNIQUENAME importance being because the default location for a port's > >> > OPTIONSFILE is derived from it, and non-uniqueness can lead to ports > >> > fighting over control of that file? Which is bad when unintentional, > >> > but can be useful for some related ports to share the same options > >> > settings. > > > Well this is the right thing to do but looking at bsd.port.mk and the > > changes > > needed I get bored and gave up :( > > I haven't looked at what would be necessary to fix UNIQUENAME collisions > in any great detail, but I think a number are due to setting PORTNAME to > something basically incorrect.
In fact there are 3 majors guilty:
- bsd.python.mk
- bsd.ruby.mk
- bsd.apache.mk
why because they recommand their user to set PKGNAMEPREFIX to
${BLA_PKGNAMEPREFIX} so that their prefix can be set from bsd.port.mk
Which mean that until the said bsd.$guilty.mk is included UNIQUENAME name is set
to ${PKGNAMEPREFIX}${PORTNAME} with PKGNAMEPREFIX being expanded to nothing
and once bsd.$guilty.mk is included ${PKGNAMEPREFIX} expands to something and so
UNIQUENAME is changed. (btw also break the options framework (old and new) with
the ports because OPTIONSFILE is created using the ${UNIQUENAME}
second problem is that uniquename can potentially be shared between ports for
example nginx and nginx-devel and in general all the -devel ports
>
> >> > Does pkgng really need LATEST_LINK at all? As far as I recall, that
> >> > only exists so that the user can say:
>
> > Well no pkgng doesn't need it at all except for pkg itself for the
> > bootstrap :)
>
> Hmmm... so there just has to be pkg.txz at some predictable URL(s) on
> the FBSD mirrors? That doesn't sound like enough to justify keeping
> LATEST_LINK related bits in the ports tree.
Once pkg_install is gone then LATEST_LINK can go away
> I think I see. You're thinking of packages that install the same files,
> but maybe in a different location (eg. SITE_PERL) or that register
> run-time dependencies on a number of different possible providers.
>
> Couldn't that boil down to having several alternate .MANIFEST files in
> the pkg? Plus some sort of final-location-independent way of naming the
> files to be installed by the package?
>
> On the matter of having alternate RUN_DEPENDS to be set at install time?
> I've wanted to do something like that with databases/phpmyadmin for
> ages. Most of the optional dependencies there are autodetected by the
> PHP code at run-time, so it should be possible to just ask the user
> which of them they want to have during pkg installation.
That is over complicated imho, and I don't like much the interactive installers,
most of people would want to automated the whole installing process.
regards,
Bapt
pgpJJzV8oR0gm.pgp
Description: PGP signature
