On Sun, 14 Apr 2013 12:25:27 +0200
Pacho Ramos <pa...@gentoo.org> wrote:

> El dom, 14-04-2013 a las 12:08 +0200, Michał Górny escribió:
> > On Sun, 14 Apr 2013 11:59:14 +0200
> > Pacho Ramos <pa...@gentoo.org> wrote:
> > 
> > > El dom, 14-04-2013 a las 11:45 +0200, Michał Górny escribió:
> > > > On Sun, 14 Apr 2013 11:40:03 +0200
> > > > Pacho Ramos <pa...@gentoo.org> wrote:
> > > > 
> > > > > # >=mono-0.92 versions using mcs -pkg:foo-sharp require shared 
> > > > > memory, so we set the
> > > > > # shared dir to ${T} so that ${T}/.wapi can be used during the 
> > > > > install process.
> > > > > export MONO_SHARED_DIR="${T}"
> > > > 
> > > > Don't use ${T} in global scope. And just don't export them
> > > > in the global scope either.
> > > 
> > > Why not?
> > 
> > Let's start with the fact that ${T} is only partially persistent
> > by the words of PMS. I don't know if it's really relevant here but
> > you're exporting persistent variables with value based on
> > an non-persistent one.
> > 
> > Thinking about it more, it probably would work. As long as you don't
> > assume anything about those directories on pkg_*rm() where ${T} would
> > have changed already and your variables wouldn't.
> > 
> 
> Yes, they will be needed at compile time, that would explain why no
> problem raised for now :/ Thanks for the info

Well, I am a bit worried about metadata regen but I guess even having
invalid values there doesn't hurt anyone. Well, assuming that the
eclass will always be sourced before running the build and noone will
try to use a pre-cached environment.

> > Also, why are you exporting HOME? PMS does that already...
> > 
> 
> Probably because it's inherited from current mono.eclass, but, are you
> sure PMS does that already? There are more examples in the tree (in
> eclasses and ebuilds) exporting HOME in similar way (vim.eclass for
> example) :/ Or maybe it was started to be exported more recently and
> this is only a relic :|

It's on the list [1], and I think I've seen 'home' subdirs in the build
root for a long time. I guess people may be resetting it 'just to be
sure'.

One thing should be noted though: $HOME may and usually will be
different than $T.

[1]:http://dev.gentoo.org/~ulm/pms/5/pms.html#x1-11800011.1

-- 
Best regards,
Michał Górny

Attachment: signature.asc
Description: PGP signature

Reply via email to