Re: [gentoo-dev] RFC: ACCEPT_LICENSE and deprecation of check_license

["Petteri Räty" <betelgeuse@gentoo.org>]

On 21.2.2010 14.49, Zac Medico wrote:
> On 02/21/2010 02:36 PM, Petteri Räty wrote:
>> On 21.2.2010 14.17, Zac Medico wrote:
>>> On 02/21/2010 09:08 AM, Petteri Räty wrote:
>>>> On 20.2.2010 14.28, Zac Medico wrote:
>>>>> Hi,
>>>>>
>>>>> Since portage-2.1.7.x is stable now, with ACCEPT_LICENSE support, we
>>>>> can think about deprecating check_license [1]. This will allow us to
>>>>> avoid using PROPERTIES=interactive in cases when it is due to
>>>>> check_license alone, since anything with a license in the @EULA
>>>>> license group is automatically masked by the default
>>>>> ACCEPT_LICENSE="* -@EULA" portage configuration [2].
>>>>>
>>>>> [1] http://bugs.gentoo.org/show_bug.cgi?id=299095
>>>>> [2] http://bugs.gentoo.org/show_bug.cgi?id=302645
>>>>
>>>> We could handle it like deprecating ebeep and epause. With EAPI=4 don't
>>>> define the function any more and the Portage version will be
>>>> sufficiently new to have ACCEPT_LICENSE.
>>>
>>> That's a good idea. However, we may want to deprecate check_license
>>> it starting with EAPI=3 since the corresponding portage versions
>>> already support ACCEPT_LICENSE.
>>
>> Likely there wouldn't be any breakage with it doing it in EAPI 3 but it
>> would be against the eclass contract of not changing expected behavior.
> 
> Given that check_license already returns silently if the user has
> accepted the appropriate license(s) via ACCEPT_LICENSE, it's not
> necessary to change the eclass contract in order to safely remove
> PROPERTIES=interactive from EAPI=3 ebuilds.

So we could keep check_license defined in EAPI 3 and remove interactive
from PROPERTIES and in EAPI 4 undefine it. We should also have a repoman
check so developers catch it.

Regards,
Petteri