Re: pkg_comp - asking for setup tips

[Robert Elz <kre%munnari.OZ.AU@localhost>]

    Date:        Sun, 12 Jul 2009 18:50:44 -0400
    From:        "Martin S. Weber" <Ephaeton%gmx.net@localhost>
    Message-ID:  <20090712225044.GA506%agamemnon.entropie.local@localhost>

  | Actually the only two thing that annoys me about 
  | pkg_comp is that PKG_DEVELOPER is set inside the chroot's etc/mk.conf

I'm not sure that's a bad thing - in fact, I'm not sure that PKG_DEVELOPER
shouldn't just go away completely as an option, with pkgsrc working as
if it is set, always.   The cost of including it seems pretty small to
me (as in, I do all my pkg building via pkg_comp, so PKG_DEVELOPER is
always set, and I'm not aware of any particular costs).   On the other
hand, as it is clear that many (perhaps most) actual pkg developers don't
bother to set it, we end up with bugs in packages that would have been
trivially fixed if the developer had used this resource.  Making it
mandatory (or at least default to on, requiring specific action to disable)
would avoid many of those issues.

  | and packages "failing" because they're built as dependency already

yes, that one is a nuisance, though it isn't pkg_comp's fault - pkgsrc
could entirely reasonably simply skip over packages when it finds the
exact same version already installed (and ideally check for that before
(re-)compiling rather than after...  (strictly: before, in addition to after,
some other parallel build may have installed it while this one is compiling).

I use a script that attempts to find and sort all out of date binary
packages, and then sort their build order to avoid this problem - if
my script was perfect (it isn't, and I haven't worked out why yet) this
problem would never be observed (by me) - the dependency would always be
built explicitly before ever being needed as a dependency, so the explicit
build would never fail.   It's just a matter of getting the sort order
right I'm sure...

kre