- Jan 22, 2018
-
-
Matthew Fernandez authored
-
Matthew Fernandez authored
-
Matthew Fernandez authored
It turns out there's a case inside guards (and possibly other contexts) where the state reference we have is const and, for this reason, we want any references to state variables to be const. We achieve this by calling an overloaded static function that allows us to return a qualified constructed object.
-
Matthew Fernandez authored
-
Matthew Fernandez authored
-
Matthew Fernandez authored
-
Matthew Fernandez authored
-
Matthew Fernandez authored
-
Matthew Fernandez authored
-
Matthew Fernandez authored
-
Matthew Fernandez authored
-
- Jan 21, 2018
-
-
Matthew Fernandez authored
We now model references to state variables and local quantities using the same classes. The type of any particular variable is determined at construction time.
-
- Jan 19, 2018
-
-
Matthew Fernandez authored
Because we constant fold them, they can easily trigger false positive compiler warnings about being unused.
-
Matthew Fernandez authored
-
Matthew Fernandez authored
There is some ambiguity here in some places, so best to be explicit.
-
Matthew Fernandez authored
My thinking is that we can emit a single type that models both a reference and a value. Which is in use will depend on the constructor called, which should always be static. As a bonus, I think this should give us functions/procedures for free without having to tweak this scheme.
-
Matthew Fernandez authored
-
Matthew Fernandez authored
-
- Jan 18, 2018
-
-
Matthew Fernandez authored
-
Matthew Fernandez authored
-
Matthew Fernandez authored
-
Matthew Fernandez authored
This is necessary to allow referencing state variables at runtime.
-
Matthew Fernandez authored
-
Matthew Fernandez authored
All the better to catch you with, my pretty.
-
Matthew Fernandez authored
This ensures we discover their non-constantness or things like division by zero at code generation time instead of checker runtime.
-
- Jan 17, 2018
-
-
Matthew Fernandez authored
Note that there are no accessors for the state generated as yet.
-
Matthew Fernandez authored
-
Matthew Fernandez authored
This relates to the previous commit and trying to unify the referents of ExprIDs. This should pay off in simplification in the long run. Note that this restores the ability to reference state variables.
-
Matthew Fernandez authored
It turns out that, in practice, the only thing a symbol can refer to is a declaration (with some minor wrinkles to deal with Enums). To simplify things, we can now simply unify ExprIDs as referring to Decls.
-
Matthew Fernandez authored
This is part of a larger change towards lookups of symbols seeking Decls, rather than something like a Var. In the short term, this breaks Enums and their lookup resolution.
-
Matthew Fernandez authored
-
Matthew Fernandez authored
Bugs Everywhere is no longer maintained and the only open bugs we had were actually already dealt with.
-
Matthew Fernandez authored
-
Matthew Fernandez authored
-
Matthew Fernandez authored
-
Matthew Fernandez authored
-
Matthew Fernandez authored
It seems we will probably not need this.
-
Matthew Fernandez authored
-
Matthew Fernandez authored
-
- Jan 16, 2018
-
-
Matthew Fernandez authored
-