Texinfo / parse-info stuff



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

    Rupert> Raymond Toy <toy.raymond at gmail.com> writes:
    >> Yes, I'm pretty happy with these changes.  Except for the fgrep.  We
    >> should probably change common-lang-recode.mk to be created via
    >> configure so we can do the right thing with fgrep (and iconv and
    >> recode).

    Rupert> Ah, you mean use $(FGREP) like $(MKDIR_P) etc.? Cool, but we don't need
    Rupert> to make common-lang-recode.mk via configure: it gets included into
    Rupert> Makefiles by Automake, so all of these variables get set and we can just
    Rupert> use them. @FOO@ doesn't work, I don't think, but $(FOO) will.

That would be perfect!  Just need to add a test in configure to see
that fgrep exists.  Or just use plain grep; it doesn't look like we're
doing anything special with fgrep.

    >> Overall, I find the the parse-info code works fine except that things
    >> like ? mnewton return two entries.  The old system returned just one.
    >> This needs to be fixed.  And we also need to test this on Windows to
    >> make sure it works there.

    Rupert> Yes, I know that still needs doing. It's because I've not yet put the
    Rupert> logic (back) in to ignore top-level chapter headings. That should be
    Rupert> easy to fix, since I put in the infrastructure it needs when I was doing
    Rupert> stuff at run-time.

Sounds like parse-info is pretty close to being equivalent to the old
method.

Oh, on my slow solaris box, I think the parse-info stuff is quite a
bit faster than the old perl method, even when the parse-info stuff is
run for each configured lisp.  I'll try to get some more precise
timing, just so we know.

Ray