search

shingarov / Pharo-ArchC

Generate binutils from formal spec
MIT License
4 stars 1 forks source link

Use `Z3Solver >> #release` instead of `#del` #37

Closed janvrany closed 2 weeks ago

janvrany commented 2 weeks ago

MA commit cceedbd removed #del in favor of #release. This commit catches up to use the latter.

janvrany commented 2 weeks ago

CI fails because 

    self assert: constituents size equals: 3.

fails. That's because encoding is BV32:(concat (concat (concat #b000011 to) #b01100) #x3039). Indeed, encoding simplify yields BV32:(concat #b000011 to #b011000011000000111001)

@shingarov Do you recall, did we change something in MA that drops that simplify somewhere?

shingarov commented 2 weeks ago

So merging #36 did NOT fix CI. Actually I would be surprised if it did. ArchC is massively broken at this point; I need to dedicate some love to it.

did we change something in MA that drops that simplify somewhere?

We absolutely did, as part of the PLE chain of commits. That simplify is a very deep rabbit hole. I believe its investigation is blocked by integrating Nikolaj's fix to "wrong arity". Which is blocked by Settings (or what was that thing called). Which, IIRC, was blocked by Ref Counting (or was it?)

janvrany commented 2 weeks ago

We absolutely did, as part of the PLE chain of commits.

Ah, of course, thanks!

janvrany commented 2 weeks ago

ArchC is massively broken at this point; I need to dedicate some love to it.

Let's delay this a a bit before having a conversation about it. There're a number of cases where the current design make things bit awkward.