[Bug 458367] Review Request: ocaml-ocamlgraph - OCaml library for arc and node graphs

[bugzilla at redhat.com]

Please do not reply directly to this email. All additional
comments should be made in the comments box of this bug.


https://bugzilla.redhat.com/show_bug.cgi?id=458367





--- Comment #3 from Alan Dunn <amdunn at gmail.com>  2008-08-07 17:56:44 EDT ---
(In reply to comment #2)
> Okay, I'll pick this one up.  Here are a few quick comments:
> 
> rpmlint produces an error which you've not justified.  When I run:
> [dwheeler at dwheeler2-pc SPECS]$ rpmlint ocaml-ocamlgraph.spec
> ../SRPMS/ocaml-ocamlgraph-0.99c-1.fc9.src.rpm ../RPMS/i386/ocaml-ocamlgraph-*
> ocaml-ocamlgraph.i386: E: no-binary
> 3 packages and 1 specfiles checked; 1 errors, 0 warnings.
>
> $ rpmls ../RPMS/i386/ocaml-ocamlgraph-0.99c-1.fc9.i386.rpm
> drwxr-xr-x  /usr/lib/ocaml/ocamlgraph
> -rw-r--r--  /usr/lib/ocaml/ocamlgraph/META
> -rw-r--r--  /usr/lib/ocaml/ocamlgraph/graph.cma
> -rw-r--r--  /usr/lib/ocaml/ocamlgraph/graph.cmi
> -rw-r--r--  /usr/share/doc/ocaml-ocamlgraph-0.99c/LICENSE
> 
> These aren't really architecture-dependent files, correct?
> Since .cma is bytecode, and .cmi is a compilation signature, these
> are NOT architecture-dependent, and they should be in /usr/share/WHATEVER...
> correct?  Let me know if I'm wrong about this, but I think
> rpmlint gets this one right, these AREN'T architecture-dependent,
> and so the FHS says that they belong in /usr/share and not /usr/lib.
> (Yes, this is an FHS rule that's frequently violated.)

(In reply to comment #1)
> ocaml-ocamlgraph.i386: E: no-binary

You just reminded me of the thing I forgot to enter into the original
description:

I'm aware of this, however, may actually not be considered an error. It is
currently standard practice with OCaml library packages to keep them as
architecture dependent packages due to the limitations of the RPM format as per
the following discussion:

https://www.redhat.com/archives/fedora-packaging/2008-August/msg00017.html

A potential fix is suggested in

https://www.redhat.com/archives/fedora-packaging/2008-August/msg00020.html

However, this is currently not what has been done for previous packages. The
question becomes whether now is the time to deviate from prior practice.

> I'm very glad to see a "%check" section.  But I think you should NOT do:
>  cp %{SOURCE1} .
> during %check; I think you should do that during %build.  I would expect the
> result of "%build" to be that "all the files are ready for checking and
> installing", and that's not really the case here.  If I built, and did some
> hand-testing, the 'expected output' file isn't in the BUILD directory yet in
> the current approach.  It's also odd that a SOURCE file is NOT used in the
> %build phase, but only afterwards.

Okay, I'll change that.

-- 
Configure bugmail: https://bugzilla.redhat.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.