Re: man3 location for ocaml

["Johnny C. Lam" <jlam%NetBSD.org@localhost>]

On Fri, Aug 11, 2017 at 05:23:45AM +0000, David Holland wrote:
> On Thu, Aug 10, 2017 at 05:20:13PM +0000, Johnny C. Lam wrote:
>  > I'm looking at the history of lang/ocaml and trying to understand
>  > why the section 3 manpages for OCaml are being installed into
>  > /usr/pkg/lib/ocaml/{share/man,man}/man3 and then symlinked into
>  > /usr/pkg/{share/man,man}/man3.  That seems convoluted for no
>  > beneficial reason.
> 
> ISTM that since they're pages about functions in a different language,
> they're not in the same namespace as the rest of man3.
> 
> This is something that nothing has ever really handled very well
> (thus e.g. tcl has its whole own directory "mann", but why "n"?)
> and it would be nice to solve for real.
> 
> Maybe man3o or man3ocaml? The latter seems better; single-letter tags
> don't scale.

man/man3/foo.3o is what OCaml installs in its default configuration.
I don't know if other packaging systems change this suffix to
something else.

> Currently the gv package installs pages "gv.3lua", "gv.3perl", and
> "gv.3tcl" in /usr/pkg/man/man3, and at least by default netbsd's
> man(1) can only address the first one unless you do man -a.

man/man3/foo.3{lua,perl,tcl} is what Debian uses now, and I think
it's a reasonable way to do it.  Presumably our manpage tools could
be adjusted to understand that naming convention as well.  Following
this convention, we should install OCaml section 3 documentation
as man/man3/foo.3ocaml.

> ... it also treats variant section names it knows about (like "3lua")
> differently from ones it doesn't (like "3tcl"), which seems like a
> problem: why doesn't it automatically discover which variant sections
> exist?

Good question.

-- 
Johnny C. Lam
jlam%NetBSD.org@localhost