- Each major module (lexer, parser, assembler) has its own error type
- These error types are somewhat interconnected, but their
dependency relationships are one-way and well defined
- The AST is no longer responsible for assembling itself
- The Assembler (assembler::Assembler) will now visit every AST node
and accumulate words
- Words are assumed to be little-endian.
- There are now a set of assembler directives that affect the
generated output:
- .word <Number>: inserts a single word in the output
- .words [<Number>,*]: inserts multiple words in the output
- .byte <Number>: Alias for .word
- .bytes [<Number>,*]: Alias for .words
- .string "String": inserts a null-terminated UTF-8 encoded string
- .strings ["String",*]: "" multiple strings
- Data is always word-aligned at the moment.
- There are now assembler directives that affect the AST during
parsing:
- .include "path/to/file": Parses the contents of a file directly
into the AST
- Included files have their own defines, but *share* labels.
This is because .defines are a tokenizer construct, and
including a file creates a new buffer and tokenizer.
- Circular includes are NOT checked for at the moment.
It is very easy to exhaust the stack.
- General cleanup of several functions, comments, TODOs, etc.
- main.rs was moved to make room for upcoming improvements to the UI
TODO:
- REPL mode is only partially compatible with .define directive
- Branching to a label will branch to the data AT the label,
not the label itself. I doubt this is correct behavior.
- In case br <label> is meant to use the absolute address,
I've created a .org directive (currently unimplemented)
for specifying the load address of the program.
