On 2016-12-18 05:34 , René J.V. Bertin wrote:
On Sunday December 18 2016 05:11:11 Joshua Root wrote:
I really can't imagine the scenario in which it's wrong. If you include
the python portgroup and have a name of the form py-something, you're
asking for this behaviour.
Maybe, maybe not. Most of the time it will work out OK and you also learn
quickly enough that this portgroup has a side-effect that you have to take into
consideration. But the principle is wrong, and there isn't even any need to do
this. To reply tit for tat: I really cannot imagine a situation in which you
*have* to reset depends_lib in a portgroup, unconditionally. Dependencies are
lists, and it's even possible to remove unwanted dependencies (a Qt5 portgroup
could remove all Qt4 possible known dependencies for instance).
Of course "you're asking for it" in a sense when you use the portgroup. That
goes for just about everything with computers, they do what you tell them to even if you
don't know what that is.
If this behaviour is required (I haven't tried to understand) it should be
documented somewhere and preferably not on line 128 ...
Are you aware that it's not just the dependencies being defined, but the
build and destroot phases too? That's what makes it a stub port.
If the documentation could be made clearer, of course I'm not opposed to
that.
- Josh
