On Jun 23, 2019, at 05:02, Mojca Miklavec wrote:
> I guess this then changes my question into: how could we get rid of
> these two rows?
>
> master_sites pypi:[string index ${python.rootname}
> 0]/${python.rootname}
> distname ${python.rootname}-${version}
>
> We still need to keep in mind that some ports might prefer to fetch
> python modules from github for whatever reason, so it needs to keep
> working in cases like those.
>
> Maybe something along the lines of "python.use_pypi yes" which would
> set the two?
I think that would be my recommendation, yes. The similar functionality in the
php-1.1 portgroup (php.pecl yes) has been in place for years and works fine as
far as I can tell (however most php modules are maintained by me, so it's
possible that if there were many others maintaining php modules, their unique
coding styles might reveal problems with this approach that I haven't
encountered with my coding style). python.use_pypi would be an option, it would
default to no, and you would add an option_proc to it (like php-1.1's
php._set_pecl) so that when it is set to yes, it would set pypi-appropriate
values for distname, master_sites, and the livecheck options. Maybe it should
only set the *defaults* of those options, rather than setting the options
themselves... but I see that in the php-1.1 portgroup I'm setting the values,
not the defaults, and I don't remember if that was intentional. Oh well.
