Do not extend nREPL with registers -- the protocol is trivial, the
implementation is a whole debugger. Emit DWARF from the LLVM backend
and use lldb-dap with dape; no adapter is written. C layout with no
object headers means DWARF describes Flan structs exactly, so lldb
needs no plugin or formatters.
Reload mostly works with DAP for free: lldb watches the loader
rendezvous and re-resolves breakpoint specs against new modules. The
gotcha is our own never-unload rule -- N redefinitions leave N copies
all claiming the same source line, so one breakpoint resolves to N
locations. Needs per-generation DWARF source identity plus a reload
agent that disables superseded locations.
Cuts the instrumentation-based step debugger, the most expensive piece
of milestone 8. Knock-on: &env stops being urgent, and the shadow stack
shrinks to serving nREPL backtraces and restart enumeration.
Second stage of the milestone-2 frontend. calc-me.flan (12 decls) and
sand.flan (20 decls) both parse end to end, and both are test deps so a
regression fails `dune test` rather than surfacing at the CLI.
Three silent-misparse bugs fixed along the way -- all cases that read
cleanly and meant something else:
- dotimes/defer/some/try/fn fell through to Call, discarding their
binding and control-flow meaning. Now special forms. Forms from later
milestones (handler-bind, restart-case, loop/recur, defmacro, signal,
with-allocator, errdefer, await) are rejected outright rather than
parsed as calls.
- (Some 1) in first body position was read as a return type, because
(Option f64) and (Some 1) are identical s-expressions and the
heuristic was capitalisation. Now decided by the set of names actually
declared as types, collected in a pre-pass -- exact, and
order-independent so a type declared below its user still resolves.
- Array literals in value position were rejected outright.
Also adds NEXT.md with the handoff for the checker.
