Re: issue with fake target with LOCALBASE

Hi,

Solene Rapenne wrote on Fri, Jul 27, 2018 at 12:48:36PM +0200:
> Solene Rapenne <solene@perso.pw> wrote:
>> Antoine Jacoutot <ajacoutot@bsdfrog.org> wrote:

>>> IMHO you are opening a can of worms. I always wondered why we even
>>> wanted to support that. Even if you get this fixed I am pretty sure
>>> you'll end up in other weird cases. I'm part of the 'let us just
>>> hardcode /usr/local and be done with it' gang. Too much magic for
>>> me. I think it's more interesting to work on having pkg_add being
>>> able to install under a different prefix.

>> I see 4 solutions about this issue
>>
>> - drop LOCALBASE support using another value than /usr/local
>> - fix building using another LOCALBASE
>> - keep it in the current state which seems to be broken
>> - as you suggested, modify pkg_add to install packages in another
>> prefix (I don't know how it could works as some applications
>> hardcode path at compile time)
>>
>> I don't know who can decide of that though.

espie@ and ajacoutot@ would seem like a good team for deciding that
to me, taking input from other porters working on infrastructure
into account.

> bsd.port.mk(5) says
>
> BUGS AND LIMITATIONS
> LOCALBASE, X11BASE, SYSCONFDIR and PREFIX are not heeded consistently.
> Most of the ports tree will probably fall apart if one tries to build/use
> stuff elsewhere.

To me, that sounds like a very strong argument to remove support
for these features as suggested by ajas 'let us just hardcode
/usr/local and be done with it' gang.

I mean, seriously, since when does OpenBSD indulge in providing bells
and whistles that are not even *supposed* to actually work? Besides,
i have often heard even core porters groan that some aspects of the
ports infrastructure are hard to master - most of that is certainly
unavoidable because porting is a complex business, but when aiming
for KISS, every small bit helps, and this does seem like an obvious
chance to get rid of some complexity.

And i can't count the number of times that i saw people told that
they mixed up LOCALBASE and PREFIX.

> so I think that we can keep things as there are.

You mean, keep a feature that doesn't fully work, isn't even supposed
to, and is almost never tested? That doesn't make sense to me, in
particular not on OpenBSD. I'm not a very prolific porter, though...

Yours,
Ingo