Dylan Baker

Rather than printing in the MIR lowering passes, emit some kind of `message()` object that all gets resolved at the end of the program. This may require allowing for a...

set_variable ========= We have the problem that you can use `set_variable` to set a variable by a variable. consider: ```meson foreach x : ['a', 'b', 'c', 'd'] set_variable(x, 'foo') endforeach...

fixing up the phis is something that basically always need to happen when two blocks are joined (assuming there are phis), so just combine them into one pass.

Especially with `continue` and `break`, it's very possible to end up in a situation where there are multiple exits from a block to the same block, as: ```meson foreach a:...

It's valid for a dictionary to have a key that is a String, Identifier pointing to a string, or some callable that returns one of the previous things. We currently...

We don't want to print the warnings, errors, and messages unless there actually is something to print.

The pass that merges them is fine, and could possibly stay. But it makes more sense to distribute the arguments at the MIR level, and not worry about them at...

The implementation in the MIR structures and FIR structures and lowering, as well as the Ninja backend exists, however, it's not exposed in a user facing way.

This should be based on the number of hardware threads we have available, which we need a utility to query. Bonus points if we spawn less than the maximum number...

Maybe something that's part of the Meson++ binary, maybe something that's not?