torque-asm/src/stages, branch v3.0.0Lightweight meta-assemblerhttps://code.benbridle.com/torque-asm/atom?h=v3.0.02025-10-15T00:25:28ZRaise an error if recursion depth exceeds a maximum value2025-10-15T00:25:28ZBen Bridleben@derelict.engineering2025-10-15T00:25:28Zurn:sha1:a4c2b3ae304dc53e8e52cc952383b5171ce00b99
This is preferable to having the assembler crash with a stack overflow
error, because the user can now see which invocation caused the
overflow.
Fix incorrect first segment address2025-10-15T00:15:27ZBen Bridleben@derelict.engineering2025-10-15T00:15:27Zurn:sha1:c62da877b370392accfbd8a1087ad5673b6f55e4
Previously, if the first intermediate token in a program was a pinned
address, the address was being discarded and address zero was being
pinned instead.
Remove string-style argument syntax from error message2025-10-14T23:57:43ZBen Bridleben@derelict.engineering2025-10-14T23:57:43Zurn:sha1:e010ea0cc5e3e8ec210d5e469ad4a41296fc4d50
The string-style argument syntax is no longer supported, it has been
replaced with a list-style syntax (using square brackets instead of
double-quotes).
Implement new bytecode stage2025-10-14T08:33:42ZBen Bridleben@derelict.engineering2025-10-14T08:33:30Zurn:sha1:d9e0c4895608cdcb04b868222b49b3f117766ed0
This completes the Torque version 3 rewrite, other than some extensive
testing that is yet to be done.
Implement new intermediate stage2025-10-14T08:05:05ZBen Bridleben@derelict.engineering2025-10-14T07:40:39Zurn:sha1:981bb70e5077bd30ef85a0092117a875dcc614fc
Massive improvement. Label references can be used anywhere in the
program, with the program being assembled repeatedly until all labels
have stabilised. The bytecode stage will just be a tiny stage tacked
onto the end, rather than the old bytecode stage that would resolve
labels and expressions.
Track more information with SymbolParser2025-10-14T03:54:44ZBen Bridleben@derelict.engineering2025-10-13T06:22:36Zurn:sha1:d95f62157bc17ee8aad12aa8887315766207c326
To be able to re-use the symbol parser code in the intermediate stage
in order to quickly find all symbol definitions and invocations in the
program (in particular now that macro definitions are indistinguishable
from label definitions, since they can both follow references), the
symbol parser now tracks additional information about each symbol
(argument count, torque-specific role, and parent macro definition).
Implement lists as a first-class type2025-10-14T03:54:44ZBen Bridleben@derelict.engineering2025-10-13T05:14:51Zurn:sha1:0f32d50293ffb1a990acf9eec128415ba54a9561
Previously, strings were implemented as lists of integers, but there
was no way to create an arbitrary list of integers to pass to a macro
invocation.
This commit extends the []-delimited expression syntax, treating
expressions that contain only integers and invocations as lists of
integers. Strings are implemented as a sub-type of list.
Change string argument syntax to list argument syntax2025-10-14T03:54:44ZBen Bridleben@derelict.engineering2025-10-13T03:02:33Zurn:sha1:a051139f6d59cbc77eff99e9417dc21d87eaba6a
Changes string arguments from "name" to [name], to prepare for generic
lists in the language. Doesn't compile.
Allow a macro invocation to precede the macro definition2025-10-14T03:54:44ZBen Bridleben@derelict.engineering2025-10-13T00:50:19Zurn:sha1:647b10dc6c2efb2bc59f64dfef07fa5a18d24f77
There is no longer any reason to enforce this restriction. This will
allow the use of more complex macro recursion by programs.
Include argument count in the signature of a symbol2025-10-14T03:54:44ZBen Bridleben@derelict.engineering2025-10-13T00:47:42Zurn:sha1:2cc7d4d11a596798f8271f5c0b8a6e8941c32fb6
This is done by suffixing the argument count to the symbol name using
the ':' separator character, which can't normally be used in a symbol.