Texinfo / parse-info stuff



>>>>> "Rupert" == Rupert Swarbrick <rswarbrick at gmail.com> writes:

    Rupert> Raymond Toy <toy.raymond at gmail.com> writes:
    >> Too bad there's no code review tool on sourceforge.

    Rupert> Well, a mailing list with developers on it isn't so bad. I guess one can
    Rupert> always paste offending hunks inline...

It makes it kind of convenient.

But anyway, here are a few more comments.

parse-info.sh is full of bashisms, so we might as well say so and use
#!/bin/bash instead of #!/bin/sh.  (This shows up on solaris where sh
is a barebones sh.)

common-lang-recode.mk uses grep -F.  This isn't supported on solaris,
but solaris has fgrep.  I guess to be portable, common-lang-recode.mk
should be created from configure and configure should check for
fgrep.  Maybe want to check for iconv and recode which I think
configure already does.


Also, the part in parse-info.sh that tries to load the compiled
parse-info could probably be done more portably using
compile-file-pathname like so:

pi_compiled=$(find "$pi_dir" -name "parse-info-$lisp*" | head -1)
if test -n "$pi_compiled"; then
    pi_lisp=$pi_dir/parse-info-$lisp.lisp
else
    pi_lisp=$pi_dir/parse-info.lisp
fi

$maxima_local --very-quiet -l $lisp <<EOF
:lisp (load (compile-file-pathname "$pi_lisp"))
:lisp (parse-info:make-info-offsets "$info_file")
EOF

Also, I think head -1 is more portable than head -n 1.

And surprisingly, gcl has compile-file-pathname so this replacement
works just fine.

With these few changes, I can build the documentation on solaris.

What do you think?

Ray