svn commit: r394970 - head/devel/love07

[Alexey Dokuchaev]

On Fri, Aug 21, 2015 at 08:24:06PM +0300, Dmitry Marakasov wrote:
> * Alexey Dokuchaev (danfe at FreeBSD.org) wrote:
> > (Picking a random commit) I don't quite get it.  Pregen'ed files like
> > changes.txt and readme.txt can be just always installed into stagedir,
> > no?
> > So a port is built+installed once, then multiple flavors of packages
> > are generated by taking needed files from the stagedir.
> > 
> > That said, `post-install-FOO-on' target should be used when FOO bits are
> > not necessarily available (e.g. because they have to be built first, and
> > that step could/should be optional).
> 
> This has nothing to do with flavors, these are targets which depend on
> whether option is enabled or disabled.

May I ask you to quote more accurately?  It's hard to follow since I did not
say anything about flavors (which, FWIW, were used purely as an example) in
the second paragraph (which was essential, yet went unreplied).

Apparently I did not make myself clear.  I understand the "targets which
depend on whether option is enabled or disabled" part; what I don't
understand is the necessity of conditional installing of handful of text
files whose existence does not depend on any option.

After we entered staging era, we started to get rid of .if ${PORTDOCS}
checks because we could use the benefit of dumping everything possible [1]
into stagedir and then cherry-pick the bits we want to package (it can be
one package or several depending on the options).  These `post-install
-DOCS-on: <install readmes and changelogs>' targets kind go against the
idea of staging as it looks to me ATM unless I'm missing something obvious
here.

./danfe

[1] to a extent when some bits actually require to be built via Doxygen or
other latex, which means you need to fetch ~1.6G distfile, which means you
probably don't want it, or at least prefer to be asked for consent first.