I changed a little bit of the code, but mainly the tests. Several of the remaining failures are actually real failures, so I need to dig through the rest carefully. A lot are failing because the C++ backend is broken.
Due to awkward module dependencies, some functions had to be moved around to accommodate this change. Two from Types have gone to EvalLiterals, and two to CompState. Everything still compiles just as before though.
The problem was that the free name could involved in an array dimension (and hence a type) of something in the PROC. When the name was then replaced in the type, CompState was not updated to have the new type, and instead kept the old type (potentially) all the way through to the backend, where it might be used for checking the bounds of an array index (against the old name taken from CompState, not the replaced name).
With my previous change to PRAGMAs, unknown pragmas would fatally fail in the lexer, so that an unknown pragma would always stop compilation, which is not good. I've changed it more towards Adam's suggestion of re-lexing and re-parsing the pragma from the parser, so we now gracefully ignore unknown pragmas again. The lexer is a bit messy, though.
This is quite a big patch, as it reworks a large pass. The three backend passes dealing with sizes stuff have now been merged into one (because the traversal order is important).
Instead of generating sizes arrays by blindly appending "_sizes", we now create nonces and store them in the csArraySizes map in CompState, which is a bit less hacky.
Added to that, we also generate constant-size arrays (e.g. for [8]) -- which are needed in case we pass the array to a PROC that has a flexible dimension -- at the top of the whole program, and use that array for every variable with that size (so if foo and bar have the same size, we use the same sizes array from the top of the program).
All the changes in this patch should be undone as soon as possible (building the list of workspaces, and how we sort out the enrollment counts -- that should be done by pulling up PAR replicator counts outside the whole PAR), but for now, it works.
It turns out that we weren't freeing workspaces, and neither was CCSP, and hence in long-running programs we were leaking memory at a crazy rate. This fix works, but really this stuff should be in CCSP (Currently, it doesn't seem to have a ProcFree function...)
This falls a bit outside the compiler's remit, but it will be the same code to print out the lexed occam without highlighting, which I plan to use for doing libraries with Tock in a bit.
