Re: [Chicken-hackers] [PATCH] Fix r4rs, , r4rs-null and r5rs-null modules

[Evan Hanson]

On 2017-10-14 22:40, Peter Bex wrote:
> OK, now I know for sure why we didn't do that before: it causes a panic..
> This has something to do with the ordering of the requirements and what
> they define in which order.  So, I've enhanced the comment a bit to
> explain and removed the extra requirement in this new patch.

Yeah, that's exactly the situation: library needs scheme and eval needs
library, so scheme can't need eval. I think that's a very low priority
issue, though, so for right now eval is just an implicit dependency of
scheme.

Anyway, I think adding entries to the "core-unit-overrides" list isn't
the right approach for these modules. I know I set that example by
putting "scheme" there, but that was left over from my attempts to work
around the aforementioned circular dependency. The overrides list is
really only necessary for modules that do something "special", such as
depending on a unit for-syntax (in fact, that's the _only_ case for
which the list is used at the moment).

So, because these rnrs modules (and scheme, currently) all load either
zero or one unit at runtime (that is to say, they don't do anything
"special"), instead of adding them to the overrides list I think we
should just add unit information to their module records and treat them
like any other module. See the attached patch.

Cheers,

Evan

0001-Add-unit-dependency-info-to-the-scheme-and-rnrs-null.patch
Description: Text Data