Problem
The parser does not accept the variadic-parameter shape args: ... in a function signature:
export fn kprintf(fmt: &str, args: ...) { ... }
export fn format_message(format: *u8, buffer: *u8, max_len: u32, args: ...): u32 { ... }Diagnostic: error[EP0001]: Expected type name at the ... token position.
Affected kernel files
- kernel/src/arch/arm64/rpi/uart.home:894 (
fn kprintf(fmt: &str, args: ...)) - kernel/src/i18n/messages.home:663 (
fn format_message(..., args: ...): u32) - (potentially more — surfaced as the agent for #46 cleared anon-struct returns)
Expected behavior
Accept args: ... (or args: ...T for typed variadics) as the final parameter of a function signature. Inside the function body, args should be usable as an iterable of un-typed (or typed-T) arguments, lowering to whatever the codegen calling convention prefers (likely a slice or platform va_args).
Notes
- This is the same shape Zig uses for
anytype-style variadics in some forms; or C's...from<stdarg.h>. The kernel uses it for printf-family functions. - If Home's design wants to avoid runtime variadics (preferring comptime tuple unpacking), the answer is: file a kernel-side issue to rewrite these as comptime-tuple. But the kernel author wrote them in
args: ...form, suggesting variadic is the intended dialect.
Tests
- Positive:
fn foo(args: ...) {}parses - Negative:
args: ...not in trailing position (e.g.fn foo(args: ..., x: u32)) yields a clean diagnostic
Acceptance
- The 2 listed kernel files clear
home-debug astpast the variadic param point. - Existing test suite green.