Re: TODO

[Gary V. Vaughan]

Bob Friesenhahn wrote:
> 
> When evaluating the direction to take for a C-based libtool, I tend to
> think of libtool being similar to `make' in that it is a rules
> processor.  The process of "configuring" libtool would be a matter of
> selecting which collection of rules applies to the current system.  I
> see that the "rules" are scripted somehow (could use /bin/sh as `make'
> does) and are easily changed, but the core libtool engine works
> identically on all platforms, and does not need to be based on
> scripting.

What an excellent idea!  Care to throw together a straw-man design?

> The main argument to use a VM for internal libtool logic
> would be to reduce code size so the libtool footprint is smaller.

Depending on the language chosen as input to the byte-code compiler, we
might also be able to come up with something that is a better fit for the
problem space.

Writing rules-based ltmain in C might take a lot longer than writing
it in a higher level language, and will probably take longer to work
the bugs out of.

I am not (yet) especially in favour of either C or an as yet undetermined
byte-compilable language, but I would definitely like to explore the
idea of replacing the creaky old ltmain.sh code :-)

Cheers,
        Gary.
-- 
Gary V. Vaughan      ())_.  address@hidden,gnu.org}
Research Scientist   ( '/   http://tkd.kicks-ass.net
GNU Hacker           / )=   http://www.gnu.org/software/libtool
Technical Author   `(_~)_   http://sources.redhat.com/autobook

signature.asc
Description: OpenPGP digital signature