Re: [gentoo-dev] last rites: =dev-lang/perl-5.12* and family

[Ian Stakenvicius <axs@gentoo.org>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

On 01/07/14 07:57 AM, Andreas K. Huettel wrote:
> 
> perl-core/Switch is now stabilized on "amd64 ppc x86", so this
> should be fixed.
> 
>> On 30/06/14 04:46 AM, Andreas K. Huettel wrote:
>>> [snip!] * As Fabian pointed out, perl-core/Switch-2.160.0
>>> should still go stable. Fine with me (but I can't read your
>>> minds about future stabilizations, and the virtual only had
>>> ~arch reverse deps).
>> 
>> There shouldn't be any need to read minds, here -- if the
>> previous stable perl had this capability, then the new stable
>> perl should too (whether that be via internal or external module
>> package).  If it doesn't, then the upgraded-new-stable-perl isn't
>> a valid replacement for the stable version it previously
>> replaced. Remember that there could be who knows what scripts on
>> end-user systems that need this functionality, and if it's
>> suddenly not available because the only thing that provides it is
>> ~arch, then they are not left in a happy state..
> 
> I'm not fully buying this argument, since it would prevent us ever
> stabilizing a package where an upstream feature is removed.
> 

But it's not removed -- it's just packaged separately, if one takes
the entirety of dev-lang/perl + perl-core/* as meaning "perl".  If the
feature was indeed just plain gone and there wasn't a module providing
that functionality, then of course stabilization shouldn't be held up
waiting for someone to write or package it, but if it -has- been
written and packaged already, I personally think it makes sense that
it be added to the stabilization list without the need for a bunch of
people specifically requesting it.

(i realize based on what is stated below, that it wasn't stabilized
for any other perl release after 5.12 and i would guess that is likely
why it wasn't considered for stabilization for 5.16 along with the
rest of the packages in
https://458122.bugs.gentoo.org/attachment.cgi?id=339264 )


> (Switch was first released with perl v5.7.3 and removed from
> v5.13.1. So there should have been lots of time to update
> dependency declarations or pull in the external package if needed
> for non-portage scripts.)

Perhaps if it makes sense to do so, the virtual/perl-*'s should be
adjusted so that they ewarn on pkg_postinst when the older perl
version is no longer being installed, to let them know they should
update their @world??  I expect this could be a templatized message,
even...  For this particular case, having the per-core package
mentioned in the p.mask helps a lot.


>> It would be nice if something like a pkgmove be done on the
>> virtual, though, upon its elimintation; anyone that installed the
>> virtual to obtain perl Switch shouldn't have to unmerge and
>> specifically emerge perl-core/Switch.  I'm not sure if pkgmove
>> itself supports this, though..
> 
> I dont think this is what pkgmove does (this would move the "no
> installed files" vdb entry of the virtual to the perl-core name...
> we would need a functionality that only touches the world file
> alone...)
> 

Probably we would want it to do more than just change @world, for
instance the in-vdb dependency list for all other installed packages
that contain the virtual should be changed to perl-core/Switch.  But
yes I agree 'pkgmove' as it stands would probably make a mess of things.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)

iF4EAREIAAYFAlOzBBIACgkQ2ugaI38ACPCSMgEAvA3CDQ514onG5JpeVn4kukXe
m/1QS6fy9HC70LWXEFUA/2tLYz+Ljz9RonsPnrK7U+mgOiRI4icbt4d58rjLZOsP
=TecJ
-----END PGP SIGNATURE-----