Updates the symbol-operand guide: x86 now uses the same plain %[fn] as
aarch64, and a 'How the portability works' note explains the mechanism
(compiler auto-injects LLVM's :c modifier for "s" operands, equivalent
to GCC :P/%P0 for x86 calls, no-op on aarch64, overridable). Drops the
stale per-arch :P guidance; checkpoint updated.
Adds a 'Symbol inputs — "s" = fn' section to docs/inline-assembly.md
(direct bl/call, portability, the export-vs-callconv linkage point) and
logs the symbol-operand + round-trip work in CHECKPOINT-ASM.
Moves docs/inline-asm-design.md -> design/inline-asm-design.md (the
internal design record now lives under design/, separate from the
user-facing docs/). Updates all links: current/CHECKPOINT-ASM.md,
current/PLAN-ASM.md, current/PLAN-EXTERN-EXPORT.md (../docs -> ../design)
and docs/inline-assembly.md (same-dir -> ../design).
Adds docs/inline-assembly.md — a how-to guide for inline assembly in the
docs/error-handling.md style: mental model, operands (inputs / value
outputs / naming + auto-naming rule), the result-type table, volatile,
clobbers, all three `-> @place` forms (write-through / read-write /
indirect-memory), multi-instruction `#string` templates, global asm +
lib-less extern, the JIT/AOT-yes vs `#run`-no execution model, a
cookbook (read-register, x86_64 syscall, divmod), and rules of thumb.
All aarch64 snippets are verified to run; x86_64 ones are labeled. The
design doc (docs/inline-asm-design.md) stays as the internal rationale;
this guide is the user-facing companion, linked from readme.md.
`asm volatile? { "tmpl", [name]? "constraint" (-> Type | = expr), …,
clobbers(.…) }` now parses into a flat-operand AsmExpr/AsmOperand (ast.zig +
parser.zig parseAsmExpr, dispatched from parsePrimary on .kw_asm). `volatile`
and `clobbers` are recognized contextually (not reserved). `-> @place`
write-through is rejected with a clear "Phase 2" parse error.
Codegen is not implemented yet (IR op + LLVM emit are Phases C–E), so lowering
bails LOUD + named via an explicit .asm_expr arm in lower/expr.zig (not the
generic unknown_expr else) — emitPlaceholder makes hasErrors() abort the build
on the message.
The new asm_expr tag forced (and got) arms in three exhaustive Node.Data
switches: sema.zig analyzeNode + findNodeAtOffset, semantic_diagnostics.zig
checkBindingNames — each recurses into template + operand payloads.
Design: adopted the operand auto-naming rule (design §II.5) — name auto-derived
from a {reg} pin, explicit [name] only when it differs or for register-class
operands, echo form rejected. Typing-stage rule; parser stores name: ?[]const u8.
Locked with examples/1640-platform-asm-parse.sx (multi-output divmod: named
operands, register pins, clobbers — parses then bails, called from main).
Also files issue 0137 (pre-existing, orthogonal: `sx run` with no `main`
segfaults via an unguarded JIT entry lookup in target.zig — not an asm bug).
zig build test green (648 corpus, 445 unit).
Two new workstreams:
- ASM: inline assembly — asm { "tmpl", "=r" -> T, "r" = expr, clobbers(.…) },
multi-return tuples; lowers via the existing llvm_api.c (no shim).
- FFI-linkage: add extern/export postfix keywords, migrate every #foreign onto
them, then purge 'foreign' from the tree (end-state invariant).
Drop current/ from .gitignore so plans + checkpoints are tracked normally
(the dir was ignored; only checkpoints had been force-added). Includes
docs/inline-asm-design.md. specs.md change left uncommitted.
Surface rename of the signed integer family: s1..s64 become i1..i64
(u1..u64, usize, isize unchanged). 'string' keeps the s-prefix arm in
name classification; width parsing moves to the i-prefix arm next to
isize.
Internal TypeId tags follow the surface (.s8/.s16/.s32/.s64 ->
.i8/.i16/.i32/.i64), as do mono-key mangle fragments (ptr_i64,
tu_i64_bool) and all display/diagnostic formatting (i{d}).
Migrated in the same sweep: stdlib + examples + issue repros + FFI C
companions (shared symbol names like ffi_id_i64), expected
stdout/stderr/ir snapshots, specs.md, readme.md, CLAUDE.md/AGENTS.md,
implementation_plan.md, docs/, issue writeups. Vendored stb_image and
historical flow state left untouched.
zig build test: 426/426; examples suite: 595/595.
attempt-2 review fixes (docs-only; contract mechanics confirmed sound):
- README + S0.2 grep-clean: 'S0 HEAD == base' / 'S0 == base' were inaccurate
(HEAD carries the docs/examples/tests diff). Reword to: production/compiler
behavior is base-equivalent — zero src/ changes, single-author output
byte-identical to base by construction — HEAD is a distinct commit, not base.
- S0.3 ledger: drop the stale '116-class corpus' FFI wording for the grounded
live count (96 entry trees / 95 active markers), matching the S0.1 count note.
No partition / manifest / examples / harness change. Gate green:
zig build + zig build test (LSP sweep 574, no crash) + run_examples (540/0);
m3te ios-sim build via main binary exit 0.
S0 of the ratified Fork C plan (zero-legacy name-resolution redesign, S0→S6).
Pure setup/documentation: NO production code change, NO behavior change.
Single-author output byte-identical to wt-stdlib-base by construction.
Deliverables under docs/fork-c/ (docs/, not current/, because current/ is
gitignored and the contract must be committed):
S0.1 — byte-baseline + commit-discipline: the committed examples/expected/*
snapshots are the single-author byte-identity reference; the zero-diff repro is
`zig build && zig build test && bash tests/run_examples.sh`. Resolver-target set
explicitly excluded + listed. Commit-classification rule: mirror | consumer-cutover | deletion.
S0.2 — E6b disposition + two-corpus partition: transitional E6b src NOT merged
(grep-clean: no resolveRegistrationSigTypeInSource / sig_registration_mode /
e6br_gate.test.zig on baseline). Harvested 0811–0829 trees + goldens (never the
src), empirically partitioned by running each through the base compiler vs the
E6b target:
- baseline-green (mirror-equivalence): 0795–0798 (merged) + 0823, 0828 — given
examples/expected/ markers, locked into the S0 baseline.
- resolver-target (known-wrong old behavior): 0811–0822, 0824–0827, 0829 + the
re-filed E6BR-5 nested-pattern regression — a listed xfail harness under
tests/resolver-target/ (manifest + TARGET goldens, NO active marker), flips
active+green at S3.9. 0811/0829 noted as old-selector-wrong on the E6b-unmerged
base; E6BR-5 subsumed by the whole-AST resolver, NOT an E6b attempt-6.
S0.3 — A–E6 reuse/delete ledger: every load-bearing A–E6 artifact mapped REUSED
(Fork C home) or DELETED/TRANSITIONAL (S3/S6 phase); E6c/d/e dropped, F/H/I/K
absorbed/superseded.
Gate over the baseline-green corpus: zig build + zig build test (LSP corpus sweep
574 files, no crash) + bash tests/run_examples.sh (540 passed, 0 failed) all exit 0.
The trace docs predated the current formatter. Corrected against the real
output (library/modules/trace.sx to_string + examples/expected/1025-errors-
trace-format.stderr):
- error-handling.md: replace the obsolete trace example ("error trace:" /
"raised error.X" / "at func (file:line)") with the real format —
"error return trace (most recent call last):" + per-frame "func at
file:line:col" + source line + caret.
- debugger.md: drop the stale "(planned)" marker on the trace formatter
(it is implemented); the tag-name table note now cites the failable-main
reporter's "unhandled error reached main: error.X" line, not a
nonexistent "raised error.X" trace line.
The interp's .trace_frame op only yields the packed value; the separate
sx_trace_push call op is executed by the interp as a foreign call via
host_ffi/dlsym, so the prior 'no sx_trace_push call runs' / 'never calls
sx_trace_push' phrasing was wrong. The packed low word is the op's
span.start (a source byte offset), not an IR instruction offset; renamed
every ir_offset/offset reference to span.start.
The compiled backend builds each trace Frame global as an LLVM named-struct
constant over the cached getFrameStructType() layout (file, line, col, func,
line_text) via LLVMConstNamedStruct -- a type-safe LLVM struct, not the sx
Frame TypeId / normal struct-emission path. Also correct the file field to
the source basename (full paths live in DWARF).
The .trace_frame op is niladic: it carries no operand and no GlobalId.
The compiled backend yields the interned Frame global's address as the
op's value (reflection.emitTraceFrame); the interpreter yields a packed
(func_id, ir_offset) as the op's value and never calls sx_trace_push
(recovered later by .trace_resolve). The sx_trace_push call is a separate
call op emitted by lower.zig at each push site, consuming the op's value.
Reword every passage that stated the old/wrong model: the niladic
invariant is about the op (not the push site emitting only one
instruction); reflection yields the op's value rather than lowering a
push; the interp returns the packed value rather than calling the foreign
sx_trace_push via host_ffi dlsym.
Name the niladic op `.trace_frame` (no `.trace_frame_push` op exists) in
the trace-path roadmap, matching the rest of the doc and src/ir/inst.zig.
Describe the `.trace_frame` arm as building/interning the Frame global and
yielding its address as the op's value; the separate sx_trace_push call is
emitted by the lowerer via normal call lowering, not by the arm itself.
Refresh the debugging architecture reference for the A7.2 relocation:
DWARF emission lives in src/backend/llvm/debug.zig (DebugInfo) and the
interned Frame / tag-name tables in src/backend/llvm/reflection.zig
(Reflection); emit_llvm.zig is the orchestrator that owns LLVMEmitter and
dispatches to them. Behavior is unchanged; only the file-and-function map,
the 'what's emitted' home, and the debugEnabled() owner are corrected.
Closes out E3's stepping-verification ladder to the extent possible
headlessly.
- Verified `sx build --target ios-sim --emit-obj` produces an
arm64-ios-simulator Mach-O that runs under `simctl spawn` and steps
in lldb (the backtrace shows a dyld_sim frame — the sim runtime).
- Verified the device-applicable .dSYM path: dsymutil collects the
DWARF, and after removing the .o lldb still resolves source via the
.dSYM.
- debug_stepping_smoke.sh gains an optional iOS-sim rung that reuses an
already-booted simulator (never boots one — single-sim policy) and
exercises the .dSYM path; skips cleanly when no sim is booted.
- docs/debugger.md: rungs 1-2 marked verified; the iOS-device rung is
documented as a manual checklist (needs hardware + get-task-allow
signing) — no compiler gap, --emit-obj + standard Apple tools suffice.
E3 is functionally complete and verified across macOS + iOS-simulator.
`sx build --emit-obj` keeps the DWARF-bearing object so a debugger can
step the binary, completing the deep-debug half of the trace story.
- --emit-obj flag + TargetConfig.emit_obj. Implies -O0 (DWARF only
emits at opt none/less); keeps the object at its link-time path
.sx-tmp/main.o so the binary's debug map resolves to it; skips the
Level-1 binary cache; reports the object path. macOS resolves via the
debug map -> .o; Linux carries DWARF in the binary. Build-flow only,
no runtime/codegen change.
- tests/debug_stepping_smoke.sh (3e rung 1; macOS, lldb, not in
run_examples): builds with --emit-obj, drives an lldb file:line
breakpoint, asserts resolution + a source-mapped backtrace. Passing —
proves the slice 1-2 DWARF drives real source-level stepping.
(Also normalizes the 253 .exit trailing newline from the 3c --update.)
Gates: zig build, zig build test, run_examples.sh -> 291 passed.
Each trace frame now shows the offending source line with a `^` caret
under the column — in the catch-handler formatter, the failable-main C
reporter, and the comptime path.
The source line is embedded at compile time as a 5th Frame field
(line_text), not read from disk at runtime: the file field is a
basename and a runtime read would add a filesystem dependency that
fails under the test harness and on locked-down targets.
- errors.lineAt(src, offset): shared helper for the whole source line.
- Frame gains line_text (mirrored in emit_llvm getFrameStructType,
trace.sx Frame, sx_trace.c SxFrame). emitTraceFrame embeds it; the
interp .trace_resolve extracts it from the source map.
- trace.sx (new spaces helper) and the C reporter render the line +
a col-aligned caret, guarded on a non-empty line_text.
Snapshots 243/244/247/253 regenerated. Gates: zig build, zig build
test, run_examples.sh -> 291 passed.
#run failures now print the same `func at file:line:col` trace as
runtime, resolved in-process via the interpreter's IR/source tables.
- Read-side context-split op `.trace_resolve` (mirror of .trace_frame),
lowered from a name-recognized `__trace_resolve_frame(u64) -> Frame`.
- emit_llvm: inttoptr the operand to *Frame + load (the value
.trace_frame stamped in).
- interp: unpack (func_id << 32 | span.start); resolve func/file from
module.functions and line/col via SourceLoc.compute over a new
source_map (setSourceMap wired at every production interp site).
- trace.sx: frame_at -> u64; to_string routes each frame through
__trace_resolve_frame, so one source works in both machines.
Compiled path behavior unchanged (243/244/247 identical; it now loads
via the op). New examples/253-comptime-trace.sx exercises the comptime
path. Gates: zig build, zig build test, run_examples.sh -> 291 passed.
Architecture spec for the debugging subsystem: error return traces
(embedded Frame table, niladic context-split push op, the thread-local
ring buffer), DWARF debug info as a debugger-only artifact, the exact
wiring (file/function map + trace and DWARF data flows), the rationale
for choosing embedded locations over PC+DWARF symbolization, the
runtime-artifacts split, and the macOS -> iOS-sim -> iOS-device stepping
verification ladder.
