Raymond Toy <toy.raymond at gmail.com> writes:
> I have not looked at your changes yet, but does this mean if I touch a
> lisp file, every single lisp file will get recompiled?
>
> While not nice, I think that you have to do that to get the correct
> sys-proclaim files. That is, you can't just recompile just one file
> and still get a correct sys-proclaim. But I could be wrong.
Yes, it does. But that's the only way to guarantee that everything is
correct. As far as I could tell from the feedback from Robert Dodier,
Dan Gildea and Richard Fateman, this is the only acceptable behaviour if
we are using sys-proclaim.lisp.
> On the other hand, if you use gcl and never build sys-proclaim except
> when binary is made, what you ship could be subtly broken compared to
> you've been testing with. Maybe because gcl with sys-proclaim elides
> some type check because you've proclaimed functions and some code was
> previously depending on the type check.
>
> While this scenario is hypothetical, it's a possibility.
Hmm, I guess so. But that argument works with any variation in how we
build. I mean, there could be some weird interaction that means a change
I make to foo.lisp breaks Maxima when built with SCL, say. We don't
require that every change is tested on every lisp implementation with
every set of build flags before it is committed (thank goodness). I'd
argue that this sort of error should be very rare and get fixed as a
regression quickly enough. Also, we don't actually use type declarations
all that much in Maxima, I don't think, which makes it even less likely
to be a problem.
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/20130320/c4993568/attachment-0001.pgp>;