Re: CVS commit: pkgsrc/mk

[Alistair Crooks <agc%pkgsrc.org@localhost>]

On Tue, Jul 22, 2008 at 01:46:22PM +0200, Joerg Sonnenberger wrote:
> On Tue, Jul 22, 2008 at 11:44:14AM +0100, Alistair Crooks wrote:
> > I've fixed this now, but I suggest you complain to the person who
> > made the previous change (Joerg) without documenting it whatsoever.
> 
> NO_WRKDIR_SYMLINK (the older name) was added in 2005 before I was even a
> developer and changed a bit later to CREATE_WRKDIR_SYMLINK based on your
> suggestion.

Originally, there was no definition. Symlinks were always created; if pkgsrc
was read-only, this operation would silently fail.

NO_WRKDIR_SYMLINK - I dislike any type of NO_* definition, because of
its syntax - one ends up having all kinds of NO_WRKDIR_SYMLINK=yes,
which is counter-intuitive.

The definition you added, CREATE_WRKDIR_SYMLINK reversed the meaning
(fine), but defaulted to the opposite (not fine). You also did not document
what you had done, and neither did you flag it - rather you left people
to discover it. (definitely not fine).

> So don't blame me for not adding text to the pkgsrc guide.

Why ever not?

Any changes you make should be documented. Again, non-negotiable.
 
> The behavior of the symlink as stated in the mail to tech-pkg directly
> violates your own reasoning in the original commits: WRKOBJDIR is not
> making the pkgsrc tree read-only. This has created confusing over the
> year I was participating in pkgsrc (I don't even bother to look for
> older examples) and the older behavior can be restored by setting the
> variable explicitly.

No, I am being consistent with historical behaviour, which pre-dates your
use of pkgsrc.

And I fail to see how being caught changing defaults without prior
discussion or subsequent notice gives you the moral high ground.

Regards,
Alistair