Robert Dodier <robert.dodier at gmail.com> writes:
> On 2013-03-18, Rupert Swarbrick <rswarbrick at gmail.com> wrote:
>> Well, my proposal is a sort of half-way house. sys-proclaim.lisp only
>> gets compiled once from an empty tree. If a developer is working in Lisp
>> with GCL (probably rather unlikely, since SLIME won't work), he or she
>> might end up with an incorrect proclaim. But then presumably the
>> developer will realise what's going on (since the code he or she is
>> working on will suddenly behave weirdly) and can delete
>> sys-proclaim.lisp to force a recompile.
>
> I'm sorry, but this is completely incorrect. We cannot design the build
> machinery on the assumption that the developer will compensate for its
> deficiencies.
At the moment, if a developer changes the ftype of a lisp function
without
(1) changing sys-proclaim.lisp by hand or deleting it
(2) checking in the result (by hand)
then the build breaks on GCL for everyone. With my proposal, the build
would break *on his/her system* until he/she deleted sys-proclaim.lisp
and had it recompiled.
Maybe I didn't make myself clear, but I think that the current state of
affairs is considerably *worse* on GCL than what I'm proposing.
>> Frankly, if someone is writing lisp code with GCL, they're just Doing It
>> Wrong. Pretty much any other lisp implementation out there is less
>> painful to develop on.
>
> We cannot play favorites, I'm afraid. If GCL is supported at all, we
> have to do it right.
Well, yes, but see above.
Rupert
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 315 bytes
Desc: not available
URL: <http://www.math.utexas.edu/pipermail/maxima/attachments/20130318/49b4ff03/attachment.pgp>;