svn commit: r340875 - head/graphics/py-pyglet

[Mathieu Arnold]

On Fri, Jan 24, 2014 at 09:34:04PM +0400, Dmitry Marakasov wrote:
> * Mathieu Arnold (mat at FreeBSD.org) wrote:
> 
> > | I don't think that matters unless ${WRKSRC} is deleted in the
> > | process. Actually, it's even better, as if for some reason some
> > | command leaves something behind (a corefile or a log) in it's current
> > | directory, it'll end up in ${WRKSRC} and to a ports' directory.
> > 
> > Well, yes, but no, you're missing the point, the idea is to have a stable
> > and consistent way of doing things, so that if anyone needs to touch any
> > port, (s)he doesn't have figure out what the current directory currently is.
> 
> One should _never_ depend on current directory unless he changes
> it in the same line. Because it's handling is (as you say) inconsistent
> among make versions, and because it's highly error prone (changing
> or reordering lines will break unrelated stuff). Also, all talk about
> this is useless unless whole portstree is converted and then keept
> consistent, which IHMO is completely useless.
> 
> > I was fine before because each line was executed in its own shell, but now,
> > the whole bloc is executed in one shell invocation.
> 
> Has this `now' happened in HEAD?
> 
> Both 9.0-RELEASE and 10.0-RELEASE (after bmake switch) behave like this:
> 
> % cat /tmp/Makefile 
> all:
> 	@pwd
> 	@cd ..; pwd
> 	@pwd
> % make -C tmp
> /tmp
> /
> /tmp
> 
> Although I remember having cd in one line affect pwd in another some
> time ago.

The thing is, the way it works depends on wether you do parallel
building or not.

$ make all
/home/mat
/home
/home/mat
$ make all -j 4
/home/mat
/home
/home

And like I said, it's policy :-)

-- 
Mathieu Arnold
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 964 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/svn-ports-all/attachments/20140125/3f22ced2/attachment.sig>