[gentoo-dev] RFC how to handle disappearing USE in (portage) dependency calculation

[gentoo-dev] RFC how to handle disappearing USE in (portage) dependency calculation

[vivo75]

Hi everybody,
  sometimes a package depend from another with a particular USE flag
turned on, example llvm-3.2 on dev-libs/udis86 +pic
Sometimes a new ebuild can change IUSE, indeed udis86-1.7-r1 removed pic
use which was present in 1.7-r0.

This RFC is to understend what we (you actually) want the packages
manager to do in this situation, as I see it there are mainly two options.

1) when consider the dependency _always_ satisfied, if the requested USE
is not in IUSE.
  this will make user life easier, since portage now barf conflicts but
the "wrong" dependency goes unnoticed and nobody will clean the ebuilds.

2) error out always, both if requested USE flag should have been enabled
or not, since it's a bug and should be fixed.
  emerge -uDavNt will not that easy but the tree is cleaner as a
consequence, also the developer are forced^Wencouraged  to look at the
reason the USE flag disappeared analizing if their package will continue
to work.

finally the depend in llvm ebuild has this form:
DEPEND="udis86? ( dev-libs/udis86[pic(+)] )"
and the diff between udis86 ebuilds is like this:
-IUSE="pic test"
+IUSE="test"

best regards,
Francesco R.

Re: [gentoo-dev] RFC how to handle disappearing USE in (portage) dependency calculation

[Ciaran McCreesh]

[-- Attachment #1: Type: text/plain, Size: 535 bytes --]

On Fri, 07 Jun 2013 13:26:22 +0200
"vivo75@gmail.com" <vivo75@gmail.com> wrote:
>   sometimes a package depend from another with a particular USE flag
> turned on, example llvm-3.2 on dev-libs/udis86 +pic
> Sometimes a new ebuild can change IUSE, indeed udis86-1.7-r1 removed
> pic use which was present in 1.7-r0.
> 
> This RFC is to understend what we (you actually) want the packages
> manager to do in this situation, as I see it there are mainly two
> options.

This is already covered by PMS.

-- 
Ciaran McCreesh

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-dev] RFC how to handle disappearing USE in (portage) dependency calculation

[Sergey Popov]

[-- Attachment #1: Type: text/plain, Size: 1551 bytes --]

07.06.2013 15:26, vivo75@gmail.com пишет:
> Hi everybody,
>   sometimes a package depend from another with a particular USE flag
> turned on, example llvm-3.2 on dev-libs/udis86 +pic
> Sometimes a new ebuild can change IUSE, indeed udis86-1.7-r1 removed pic
> use which was present in 1.7-r0.
> 
> This RFC is to understend what we (you actually) want the packages
> manager to do in this situation, as I see it there are mainly two options.
> 
> 1) when consider the dependency _always_ satisfied, if the requested USE
> is not in IUSE.
>   this will make user life easier, since portage now barf conflicts but
> the "wrong" dependency goes unnoticed and nobody will clean the ebuilds.
> 
> 2) error out always, both if requested USE flag should have been enabled
> or not, since it's a bug and should be fixed.
>   emerge -uDavNt will not that easy but the tree is cleaner as a
> consequence, also the developer are forced^Wencouraged  to look at the
> reason the USE flag disappeared analizing if their package will continue
> to work.
> 
> finally the depend in llvm ebuild has this form:
> DEPEND="udis86? ( dev-libs/udis86[pic(+)] )"
> and the diff between udis86 ebuilds is like this:
> -IUSE="pic test"
> +IUSE="test"

What's the question here? How to handle this? Read about USEDEP_DEFAULTS
in PMS.

If you see broken packages(somebody forgot to change dependency) - file
a bug about it.

-- 
Best regards, Sergey Popov
Gentoo developer
Gentoo Desktop Effects project lead
Gentoo Qt project lead


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 555 bytes --]

Re: [gentoo-dev] RFC how to handle disappearing USE in (portage) dependency calculation

[vivo75]

On 06/07/13 14:37, Sergey Popov wrote:
> 07.06.2013 15:26, vivo75@gmail.com пишет:
>> Hi everybody,
>>   sometimes a package depend from another with a particular USE flag
>> turned on, example llvm-3.2 on dev-libs/udis86 +pic
>> Sometimes a new ebuild can change IUSE, indeed udis86-1.7-r1 removed pic
>> use which was present in 1.7-r0.
>>
>> This RFC is to understend what we (you actually) want the packages
>> manager to do in this situation, as I see it there are mainly two options.
>>
>> 1) when consider the dependency _always_ satisfied, if the requested USE
>> is not in IUSE.
>>   this will make user life easier, since portage now barf conflicts but
>> the "wrong" dependency goes unnoticed and nobody will clean the ebuilds.
>>
>> 2) error out always, both if requested USE flag should have been enabled
>> or not, since it's a bug and should be fixed.
>>   emerge -uDavNt will not that easy but the tree is cleaner as a
>> consequence, also the developer are forced^Wencouraged  to look at the
>> reason the USE flag disappeared analizing if their package will continue
>> to work.
>>
>> finally the depend in llvm ebuild has this form:
>> DEPEND="udis86? ( dev-libs/udis86[pic(+)] )"
>> and the diff between udis86 ebuilds is like this:
>> -IUSE="pic test"
>> +IUSE="test"
> What's the question here? How to handle this? Read about USEDEP_DEFAULTS
> in PMS.
yes "how to handle this", thanks for the pointer to USE-DEP-DEFAULTS,
Ciaranm you too
>
> If you see broken packages(somebody forgot to change dependency) - file
> a bug about it.
>
According to the now readed fine manual, the ebuilds seem to be right,
llvm is EAPI=5, udis86-1.7-r1 is EAPI=4 and the previous one is EAPI=3,
dependency should be considered enabled if absent.
So the bug would be in portage, but the overzealous dep resolution seem
to arise only with "--with-bdeps=y" so I'm not opening it a bug.

Re: [gentoo-dev] RFC how to handle disappearing USE in (portage) dependency calculation

[Zac Medico]

On 06/07/2013 07:16 AM, vivo75@gmail.com wrote:
> According to the now readed fine manual, the ebuilds seem to be right,
> llvm is EAPI=5, udis86-1.7-r1 is EAPI=4 and the previous one is EAPI=3,
> dependency should be considered enabled if absent.
> So the bug would be in portage, but the overzealous dep resolution seem
> to arise only with "--with-bdeps=y" so I'm not opening it a bug.

It sounds as though you've got pic in use.mask, because portage has 
different handling for masked flags prior to EAPI 5, as described here:

   https://bugs.gentoo.org/show_bug.cgi?id=442830#c3
-- 
Thanks,
Zac