Re: Libtool itself -- speeding it up -- implementation question.

[Bruce Korb]

Rob Browning wrote:
> 
> I know that libtool is written in /bin/sh, and I realize that's for
> portability reasons, but I also seem to recall some discussion of
> rewriting parts of libtool in C to speed up some of the slower
> operations.

Probably the slowest operation in this regard is getting time
to make progress.

> If this is is in fact under consideration, it made me wonder if there
> might be alternatives to C.  In particular, I wondered if it might be
> possbile to consider re-writing libtool (or perhaps parts) in perl,
> and if this would provide any substantial performance/maintainability
> benefits.
> 
> The perl core seems likely to be very widely available, and it doesn't
> appear to use libtool itself, so there shouldn't be any bootstrapping
> issues for new architectures.
> 
> Dumb idea?

Well, once the data on what needs doing is tabularized,
I'll be able to generate shell scripts, C code or Perl.
Just a matter of twiddling a template.  My guess would
be that Perl won't be a whole lot simpler than C to 
maintain.  The hard part is going to be extracting the
tabular data from the monolithic shell script.  Ick.
But I did it once before:

  http://autogen.sourceforge.net/fixinc.html

Anyway, once the data are tabularized, we can discuss this
some more.