Re: Issues with gendocs and images

[Karl Berry]

    Really, wouldn't it be saner to build in the output zone, and
    kill it on error instead?

Seems like way more trouble to me.  Anyway, as you know, I'm not about
to rewrite the whole thing.

    Here is my proposal.  It works for Bison.  

Ok, accepted for the sake of getting on with life.  In principle I would
rather rewrite the whole script in Perl than to insert Perl code for
this one subjob, but finding the time to do that seems unlikely :(.

I tinkered with the formatting, etc.  I also made a second commit
removing the PostScript output.  Let me know if I broke something.

It looks to me like if someone does an explicit <img src="http://...";>
inside @html, it'll fail.  Guess we'll see if anyone cares about that.

    I'm surprised that the tarballs do not expand in a single directory,
    but in many files.

This script did not come into being by thinking about "what is the
cleanest way to build documentation for a web site", but rather "the
webmasters are doing the same job to build manuals for multiple
packages, let's write a quick hack to automate it".  Some 20 years ago.

And I didn't write it, I inherited it.  I find it very painful to work
on at this point, so many variables and duplicated code paths have been
introduced.

    Also, it might not be so great to simply gzip the monolithic HTML
    file, as it prevents providing more files (say images for instance :).
    A tarball would make sense too.

Yeah, but in most manuals, there are no images and a single html is
enough.  I'm not enthused at the idea of super-generalizing everything
for the most complex case.  Supporting a tarball for monolithic html
when needed shouldn't be too hard, in principle ... maybe ...

Thanks,
k