* Implement a '#' prefix for raw identifiers that may alias keywords
* Review comments
* Disallow hashless raw identifiers in interpolations
* Run clang-format
* Implement custom generic tagged union `Either`
This should be more efficient than `std::variant`, while still
keeping runtime safety as it `assert`s when `get`ting values.
* Use `Either` for RPN expressions
* Use `Either` for file stack node data
* Use `Either` for `File` buffer
* Use `Either` for `STRFMT` args
* Use `Either` for RGBLINK symbol values
* Support an equivalent of `std::monostate` for `Either`
* Use `Either` for lexer tokens
* Use `Either` for symbol values
* Use `Either` for lexer mmap/buffer state
This line was missing; in its absence, as soon as the array reallocs,
all registered symbol pointers are invalidated, leading to a crash
(e.g. when generating a sym file).
Made the reallocation check into a hard error, too.
Since fileSymbols gets registered with each section, we would have to
change that, and then it cascades throughout all of RGBLINK.
This will be handled correctly in the Rust port.
Passing a count parameter to the vector constructor does not reserve
count elements but instead fills the vector with count elements.
This caused the endianness check to fail first because the first 256
characters were null bytes.
The function only stopped at the end of the *arrays*,
not of *their used portions*!
This could cause false negatives one way or the other.
This appears not to affect the palette packing, since the packing algorithm
deals with them efficiently; but it should speed up processing slightly,
and as the test changes show, it also improves the UX of palette packing
error messages!
* Clarify the operator relationship of ordinary and fixed-point numbers
* Attempt to clarify description of fixed-point numbers
* Note that RGBASM does not check fixed-point precisions
* Simplify sine table example a bit
* Remove misleading equations describing `DIV`, `MUL`, and `FMOD`
* Various minor style and formatting fixups
The generated `result.pal` can change between platforms, it turns out
(due to "unstable sort" differences), and creates additional churn if
changing anything related to palette generation.
tl;dr this makes the test less flaky in the long run.
