[gentoo-portage-dev] Problems with the new "no downgrades"

[gentoo-portage-dev] Problems with the new "no downgrades"

[Vlastimil Babka]

*portage-2.1.5_rc1 (04 Apr 2008)

   04 Apr 2008; Zac Medico <zmedico@gentoo.org> +portage-2.1.5_rc1.ebuild:
   2.1.5_rc1 release. In the event that a previously installed package has
   since been masked, emerge will no longer perform an automatic downgrade
   as part of a "world" update. You should either unmask such packages or
   else explicitly re-merge them in order to have them dowgraded to an
   unmasked version. Bug #216231 tracks all bugs fixed since 2.1.4.x.

Assuming it's because of bug 197810, but that only talks about packages 
masked by corruption. But is it really so good to apply this also to 
keyword/package.mask or even ebuild being removed?

For example, we had swt-3.3.1.1 in SLOT="3" and released swt-3.4_pre6 
with SLOT="3". Later realized it's not backwards compatible enough and 
released swt-3.4_pre6-r1 in SLOT="3.4" removing the 3.4_pre6 ebuild. So 
I would expect the slot 3 to downgrade back to 3.3.1.1 (especially if 
something pulls slot 3 via slot dep). (Note that we can't use slotmove 
because changing slot in java package means also changing where it's 
installed and expected.) Now thanks to this change, downgrade won't 
happen. I think it's not good.

VB
-- 
gentoo-portage-dev@lists.gentoo.org mailing list

[gentoo-portage-dev] Re: Problems with the new "no downgrades"

[Petteri Räty]

[-- Attachment #1: Type: text/plain, Size: 1381 bytes --]

Vlastimil Babka kirjoitti:
> *portage-2.1.5_rc1 (04 Apr 2008)
> 
>   04 Apr 2008; Zac Medico <zmedico@gentoo.org> +portage-2.1.5_rc1.ebuild:
>   2.1.5_rc1 release. In the event that a previously installed package has
>   since been masked, emerge will no longer perform an automatic downgrade
>   as part of a "world" update. You should either unmask such packages or
>   else explicitly re-merge them in order to have them dowgraded to an
>   unmasked version. Bug #216231 tracks all bugs fixed since 2.1.4.x.
> 
> Assuming it's because of bug 197810, but that only talks about packages 
> masked by corruption. But is it really so good to apply this also to 
> keyword/package.mask or even ebuild being removed?
> 
> For example, we had swt-3.3.1.1 in SLOT="3" and released swt-3.4_pre6 
> with SLOT="3". Later realized it's not backwards compatible enough and 
> released swt-3.4_pre6-r1 in SLOT="3.4" removing the 3.4_pre6 ebuild. So 
> I would expect the slot 3 to downgrade back to 3.3.1.1 (especially if 
> something pulls slot 3 via slot dep). (Note that we can't use slotmove 
> because changing slot in java package means also changing where it's 
> installed and expected.) Now thanks to this change, downgrade won't 
> happen. I think it's not good.
> 
> VB

You can use atoms like <dev-java/swt-3.4_alpha:3 to force it

Regards,
Petteri


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 252 bytes --]

Re: [gentoo-portage-dev] Re: Problems with the new "no downgrades"

[Vlastimil Babka]

Petteri Räty wrote:
> Vlastimil Babka kirjoitti:
>> *portage-2.1.5_rc1 (04 Apr 2008)
>>
>>   04 Apr 2008; Zac Medico <zmedico@gentoo.org> +portage-2.1.5_rc1.ebuild:
>>   2.1.5_rc1 release. In the event that a previously installed package has
>>   since been masked, emerge will no longer perform an automatic downgrade
>>   as part of a "world" update. You should either unmask such packages or
>>   else explicitly re-merge them in order to have them dowgraded to an
>>   unmasked version. Bug #216231 tracks all bugs fixed since 2.1.4.x.
>>
>> Assuming it's because of bug 197810, but that only talks about 
>> packages masked by corruption. But is it really so good to apply this 
>> also to keyword/package.mask or even ebuild being removed?
>>
>> For example, we had swt-3.3.1.1 in SLOT="3" and released swt-3.4_pre6 
>> with SLOT="3". Later realized it's not backwards compatible enough and 
>> released swt-3.4_pre6-r1 in SLOT="3.4" removing the 3.4_pre6 ebuild. 
>> So I would expect the slot 3 to downgrade back to 3.3.1.1 (especially 
>> if something pulls slot 3 via slot dep). (Note that we can't use 
>> slotmove because changing slot in java package means also changing 
>> where it's installed and expected.) Now thanks to this change, 
>> downgrade won't happen. I think it's not good.
>>
>> VB
> 
> You can use atoms like <dev-java/swt-3.4_alpha:3 to force it

OK that solves my problem, thanks.
But in general case I think it's still wrong. Package is found to be 
broken, gets p.masked, but people will keep the masked version and not 
downgrade. And because it doesn't even warn about that fact, they won't 
even know!

Caster

-- 
gentoo-portage-dev@lists.gentoo.org mailing list

Re: [gentoo-portage-dev] Problems with the new "no downgrades"

[Zac Medico]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Vlastimil Babka wrote:
> *portage-2.1.5_rc1 (04 Apr 2008)
> 
>   04 Apr 2008; Zac Medico <zmedico@gentoo.org> +portage-2.1.5_rc1.ebuild:
>   2.1.5_rc1 release. In the event that a previously installed package has
>   since been masked, emerge will no longer perform an automatic downgrade
>   as part of a "world" update. You should either unmask such packages or
>   else explicitly re-merge them in order to have them dowgraded to an
>   unmasked version. Bug #216231 tracks all bugs fixed since 2.1.4.x.
> 
> Assuming it's because of bug 197810, but that only talks about packages
> masked by corruption. But is it really so good to apply this also to
> keyword/package.mask or even ebuild being removed?
> 
> For example, we had swt-3.3.1.1 in SLOT="3" and released swt-3.4_pre6
> with SLOT="3". Later realized it's not backwards compatible enough and
> released swt-3.4_pre6-r1 in SLOT="3.4" removing the 3.4_pre6 ebuild. So
> I would expect the slot 3 to downgrade back to 3.3.1.1 (especially if
> something pulls slot 3 via slot dep). (Note that we can't use slotmove
> because changing slot in java package means also changing where it's
> installed and expected.) Now thanks to this change, downgrade won't
> happen. I think it's not good.
> 
> VB

Some others were complaining about this in #gentoo-dev and now what
I want to do is revert the behavior so that it's more like it used
to be. The "masked by corruption" case from bug 197810 is special
(the installed package is not actually masked) and it will be
handled without changing the behavior in other cases.

Zac
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.9 (GNU/Linux)

iEYEARECAAYFAkf7mzMACgkQ/ejvha5XGaM9MwCglI1FIn/DfixjFsiz8uy97XsM
LJ8AoJmgn4YZbt4vcdQ51G/PkUdDHM7u
=CbCl
-----END PGP SIGNATURE-----
-- 
gentoo-portage-dev@lists.gentoo.org mailing list