On 10/04/2015 02:18 AM, Alexis Ballier wrote:
> On Sat, 3 Oct 2015 12:02:02 -0700
> Zac Medico <[email protected]> wrote:
> 
>> On 10/03/2015 02:38 AM, Alexis Ballier wrote:
>>> On Fri, 2 Oct 2015 16:54:30 -0700
>>> Zac Medico <[email protected]> wrote:
>>>
>>>> On 10/02/2015 04:40 PM, Alexis Ballier wrote:
>>>>> On Fri, 2 Oct 2015 13:08:29 -0700
>>>>> Zac Medico <[email protected]> wrote:
>>>>>
>>>>>> On 10/02/2015 07:49 AM, Mike Gilbert wrote:
>>>>>>> Hello,
>>>>>>>
>>>>>>> I am getting the output below when I run repoman full for
>>>>>>> sys-apps/systemd.
>>>>>>>
>>>>>>> It is basically telling me that systemd (which is masked in the
>>>>>>> selinux profiles) cannot depend on sys-apps/dbus[systemd],
>>>>>>> because the systemd use flag is also masked.
>>>>>>>
>>>>>>> That's perfectly fine and I suppose it is valid, but there is
>>>>>>> nothing I can do to resolve it and I don't need to be reminded
>>>>>>> of it every time I run repoman.
>>>>>>>
>>>>>>> Does anyone find dependency.badmasked useful?
>>>>>>
>>>>>> Possibly, if I wanted to see dependency issues for masked
>>>>>> packages.
>>>>>
>>>>>
>>>>> why not also ignore *use.mask along with package.mask for this
>>>>> check ?
>>>>>
>>>>
>>>> Can you give a concrete example? I'm having a hard time thinking
>>>> up a reason to ignore use.mask.
>>>
>>> Well, ignoring completely use.mask won't work: people use it because
>>> the dep doesnt work and thus has missing keywords.
>>>
>>> But, maybe something in between could work: drop
>>> dependency.badmasked warnings that are satisfied when ignoring
>>> use.mask.
>>
>> Yeah, I guess that might work as an alternative to suppressing all
>> dependency.badmasked messages by default. We would need another option
>> to enable such warnings.
>>
>> Introducing special cases for use.mask/use.force like this is not as
>> simple as it might seem. If we simply discard use.force and use.mask,
>> then it can trigger other kinds of warnings. For example, consider a
>> dependency like this:
>>
>> !hardened? ( sys-apps/systemd )
>>
>> If we were to discard hardened from use.force, then repoman will show
>> an error for this dependency being unsatisfied on hardened profiles.
>> We get analogous problems when we discard flags from use.mask.
> 
> 
> What I meant is: Generate a first dependency.badmasked list like it is
> done currently. Then filter the result by ignoring/removing those that
> are satisfied without use.mask & friends.
> 
> That is:
> DEPEND="
> !hardened? ( sys-apps/systemd )
>> =sys-apps/dbus-1.6.8-r1:0[systemd]
> "
> 
> will currently generate a dependency.badmasked list on hardened profiles
> like:
> [ ">=sys-apps/dbus-1.6.8-r1:0[systemd]" ]
> 
> since '>=sys-apps/dbus-1.6.8-r1:0[systemd]' is satisfied when discarding
> use.mask, the returned list will be empty.
> 
> 
> 
> In the end, it is just an attempt at removing false-positives from
> dependency.badmasked.

That seems reasonable.

> 
> 
> 
>>> Is there anything I'm missing ?
>>
>> Maybe it's better to keep things a little simpler, and just suppress
>> all dependency.badmasked messages by default.
> 
> 
> I also like those warnings :)
> 

If we reduce the noise level as you suggest, then maybe we can keep
dependency.badmasked messages enabled by default.
-- 
Thanks,
Zac

Reply via email to