RCS needs soelim(1) and pic(1) for ‘tarball’

[Thien-Thi Nguyen]

These programs are from package ‘groff’ on my machine, and are used in
the ‘tarball’ derivation to build the manpages.  They are not needed in
the ‘build’ derivation, although their presence cannot hurt.

IIUC, this indicates the the ‘tarball’ stanza of the RCS release.nix
expression be enhanced from:

  buildInputs = with pkgs; [ automake111x autoconf ed texinfo emacs];

to something like:

  buildInputs = with pkgs; [ automake111x autoconf ed texinfo emacs
                             groff ];

(presuming ‘groff’ is enough to make those programs available).  If
so, would you please make this change (or let me know how i can effect
it, myself -- i hate bothering you for such simple stuff)?  If not,
what am i missing?  Also, would it be wise to do likewise for the
‘build’ derivation?

FYI, the present ‘tarball’ derivation succeeds despite this lack, but
only because there is no check for the output of soelim (and pic).
For example, see:

  http://hydra.nixos.org/build/851552/log
  (around line 25229, "building ci.1")

Once the ‘buildInputs’ is augmented, i will add a sanity check for
this condition, namely:

  --- man/Makefile.am   2011-01-07 17:54:46.000000000 +0100
  +++ man/Makefile.am.NEW       2011-01-07 17:54:38.000000000 +0100
  @@ -37,11 +37,16 @@
   soelim = $(SOELIM) -I. -I'$(srcdir)'
   nolf = sed '/^\.lf /d'
   
  +# We do the sanity (test -s) check at generation time since
  +# it is a maintainer concern not a user (make check) concern.
  +
   .5in.5:
        $(soelim) $< | $(PIC) $(PICFLAGS) | $(nolf) > $@
  +     test -s $@
   
   .1in.1:
        $(soelim) $< | $(nolf) > $@
  +     test -s $@
   
   $(dist_man_MANS) $(dist_noinst_MANS): REL $(bits)
   
The reason i haven't added this yet is so that downstream ‘build’
derivations can go forward, exposing debugging opportunities there
while i wait for enhanced ‘buildInputs’ in ‘tarball’.  A little
latency hiding, you understand...