Re: [Sbcl-devel] More host's gensyms

[Douglas K. <do...@go...>]

There are other constructs which are more terrifying to me than gensym.
e.g. anywhere in src/compiler/dump or src/compiler/ir1util that use
TYPECASE, it's probably using the wrong TYPEP - the CL: one, for testing
target objects. How can we deal with that?
I'm not saying gensyms aren't important, but I don't see how using _our_
gensym can ever be the wrong thing, but I can see clear ways that using the
host's TYPEP could be wrong.


On Wed, Jan 20, 2021 at 12:21 PM Christophe Rhodes <cs...@ca...> wrote:

> Manuel Giraud <ma...@le...> writes:
>
> > Christophe Rhodes <cs...@ca...> writes:
> >
> >> I favour trying to fix most of the issues to get us to a baseline of
> >> (ideally) all xc files equal between host compilers, or at least a small
> >> number of known not-equal ones, and then making systematic changes like
> >> this.  It's just too easy to regress otherwise.  Once we're done or
> >> close to done, it might be more obvious which of the cl/xc gensyms is
> >> most useful.
> >
> > Ok, I'll try this step by step approach then. The gensym shadowing looks
> > promising though: just by adding "GENSYM" to *shadows* in
> > set-up-cold-packages, sbcl/clisp goes from 97 different files to 44 and
> > sbcl/ecl goes from 115 different files to 88. But I guess that this kind
> > of modifications needs heavy testing.
>
> Well!  That is quite a big difference.  Is it a strict subset, or do any
> files start picking up differences that didn't before?
>
> Cheers,
>
> Christophe
>
>
> _______________________________________________
> Sbcl-devel mailing list
> Sbc...@li...
> https://lists.sourceforge.net/lists/listinfo/sbcl-devel
>