Implements `docs/journey-stage1b2-lsp-guidance-framing.md` (approved at
revision 4, after three review rounds). Lua, tests and docs; no Rust
change and no protocol change.
`COHERENCE.md` §1.2's canonical silence: a preconfigured server that is
not installed failed with no status message, no record and no modeline
marker, while tree-sitter highlighting kept working and masked it. Now
the status line names the command, the language and the errno; the
modeline reads `LSP:!` instead of nothing; and `M-x lsp.status` renders
a durable `*lsp*` panel.
Half of this was already built. `status_buffer_text()` and
`last_error()` have existed since M4.8, exposed to Lua and tested, with
no production caller and no buffer to render into — several doc
comments already referred to "the `*lsp*` buffer" as though it existed.
The reporting shape was likewise already adopted twice inside
`lsp.lua`; the canonical case was silent because nobody had converted
it.
Three tables with three lifetimes, because one cannot do the job:
`reported` is never cleared and includes the command, so repointing at
another missing executable reports again; `failures` is cleared by a
successful spawn so the panel goes quiet on recovery; and a
buffer-keyed projection feeds the modeline, because that provider runs
for every window on every paint and deriving an affinity key inside it
would invoke root resolvers during painting.
The memo is on the report, not the failure: the spawn is still
attempted on every file open, so installing the binary mid-session
recovers with nothing to invalidate.
Adds `tests/lsp_spawn_guidance_acceptance.rs` (16 pins) and a step-6
row to the journey ratchet.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Lv428Fth9LRtffwJSsqH7T
All four findings were the same shape: a failure that left state wrong
and told nobody.
**Delete refusals reach the user.** `reconcile_delete_and_fire` returned
`kept_modified` and `refused` and both production callers discarded
them, so a last-buffer refusal or the asynchronous modified-buffer race
left the file gone and the buffer still bound to it — and the next
`C-x C-s` recreates the deleted file. Reporting now happens inside the
shared seam, not at its call sites, for the same reason the
reconciliation does: a caller that has to remember to report is a caller
that will forget. The message names the buffers (capped, with a count
for the rest) and states the consequence, and it is written to
`EditorCore::status`, not `pmacs.error` — that channel is defined only
by a test stub, so a report there would be the same silence.
`reconcile_delete` now prefixes `kill_buffer`'s reason with the buffer
name, because "cannot kill the last remaining buffer" does not say which
buffer is now bound to a deleted path.
**The LSP subscribers stop swallowing their own failures.** Ignored
`pcall`s around `did_close`, `forget_uri`, `did_open` and overlay
re-rooting made the callback return successfully, so the
`all-must-succeed` logger had nothing to log — concretely, a stale server
made `forget_uri` raise while the callback carried on with the old
stores, routes and `documents` entry all live. A shared failure sink
attributes each step, reports on both channels, and raises **after** the
loop, so one unreachable server cannot leave every other attachment
unreconciled.
**`forget_uri` abandons requests through the established path.** It
purged `pending_routes` and `pending_external` but not the same ids
`send_request` put in `LspClient.pending`, and recorded nothing in
`cancelled_rids`. The per-rid work is extracted from
`drain_cancelled_externals` as `abandon_request` and reused, rather than
a second incomplete copy: route, client pending, cancelled record and
`$/cancelRequest` now happen together.
**Acceptance 35 is pinned.** With a plain delete the forbidden fallback
was unobservable — `find_or_open` raises out of `load_file` and the
`pcall` swallows it — so both assertions passed with the fallback
present. The plan now deletes the origin's file and recreates it, which
gives the fallback something to open and makes "restores nothing"
falsifiable. The corrected G1 explanation also reaches the production
comments, which still repeated the false `resolve_target_buffer::NotFound`
story.
New pins: acceptance 53 and 53b assert the status channel; a stale-server
row asserts attribution on both channels *and* that the healthy
attachment still reconciles; an `lsp.rs` unit test asserts the client-side
abandonment with an unrelated request as its control.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Lv428Fth9LRtffwJSsqH7T
`resource.renamed` and `resource.deleted` are declared
`all-must-succeed`, so one raising subscriber does not stop the rest
from reconciling.
`lsp.lua` gains the two subscribers. Rename runs the ordered teardown
per attachment — flush the pending didChange, didClose the old URI,
`forget_uri` against the OLD server, re-run `ensure_server` (a rename
across project roots needs a different one), didOpen the new URI, then
re-root the diagnostic overlays. Delete tears the attachment down,
because the buffer may be gone entirely and a retained record is a
dangling handle.
The workspace-edit applier captures the origin BUFFER instead of its
path, and restores nothing when that buffer is gone. A captured Lua
local is unreachable to any transaction, and the old path fallback is
what materialized a phantom empty buffer at the renamed-away path.
`fs.lua` states the overlapping-mutation serialization precondition as
a correctness rule, with the counterexample showing why no static
ordering rule substitutes for it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Lv428Fth9LRtffwJSsqH7T
Normalize batch dependency paths through the registry's lexical
canonical form so equivalent URI spellings do not revive the
initial-state preflight bug.
Separate execution-started state from the count of completed plan
items. Preflight failures retain the no-mutation guarantee, while
runtime failures conservatively acknowledge that the failing item may
itself have changed a buffer or the filesystem.
Add real-server-pump acceptance for dot-path dependency aliases,
partial text edits within one item, and resource-operation side
effects, and record the review-round corrections in the framing.
Four findings, all reproduced by the reviewer, all accepted. Two of
them are one defect class — a guard whose scope was REASONED ABOUT
rather than enumerated — so both are recorded in the framing's new §9
together with a sweep of every other place this lane decides something
is "affected".
P1 — the delete preflight broke ordered resource operations
(§9.3). Every delete was judged against the filesystem's INITIAL
state, at plan-construction time, so a valid `create X -> delete X`
was refused with a fabricated NotFound about a path the batch was
about to create; likewise `rename A -> B -> delete B`. This was a
regression this lane introduced, not a pre-existing defect.
Decision: DEFER, do not simulate. A delete whose target is related by
component-aware path containment to a path an EARLIER op in the same
plan creates, renames onto, renames away from, or removes is not
judged at plan time; the primitive judges it when it runs. Q#RD3
already calls this check a filter, not a transaction, so declining to
judge is inside its contract and refusing a legal batch is not.
Simulating instead would mean modelling filesystem presence AND the
registry's path bindings across create/rename/edit — the transaction
Q#RD3 declines to build — and a wrong simulation emits false `clear`
verdicts, which is the dangerous direction. `edit` ops are excluded
from the deferral set on purpose: an edit changes no path's existence,
so it can only turn a plan-time `clear` into a primitive-time
refusal, which Q#RD3 already documents and accepts. The
buffer-and-filesystem half therefore still fires early for any target
no prior op touches, which is what criterion 11c pins.
P1 — the production-boundary acceptances are landed (§9.5).
Criteria 11, 11a-11d, 12 (both directions), 13 and 15 now drive a real
`pmacs_fake_lsp` child over a real transport. One parameterized mode,
`applyeditplan`, replaces the eight the framing named: it reads its
whole WorkspaceEdit from a test-written file and publishes the
client's response to a sink, so each of the eight fixtures sits next
to the assertions that depend on it instead of being mirrored across
two files. Fail-closed — an unreadable plan sends no applyEdit and
reports itself through the sink, so a broken fixture cannot read as a
pass — and the sink is written `.part`-then-rename so a polling reader
never sees a partial record. There is no skip-and-return-ok arm
anywhere: `fake_lsp_path` resolves through `env!("CARGO_BIN_EXE_...")`,
a compile-time constant, so a missing binary is a build failure.
P1 — mid-batch failures were misreported as complete aborts (§9.4).
`apply_workspace_edit` now returns `nil, message, applied_op_count`,
and ONE renderer serves both the user-facing status line and the
server's `failureReason`, so the two cannot disagree. All three
callers are updated, not only the rename one.
P2 — non-recursive deletes inspected descendants (§9.2). `recursive`
is now a parameter of the shared query and descendant matching is
reserved for recursive deletes. The old doc comment argued at length
for the wrong behaviour and is replaced by the counterexample that
falsifies it: a modified buffer at `tree/gone.rs` whose file is
already gone blocked a non-recursive delete of the now-EMPTY `tree/`,
an op that would have succeeded and that removes none of that
buffer's contents. This narrows the Q#RD6 query #171 adopts.
Criterion 3's stated bite: fixed by fixing the SETUP, not the doc.
The first commit's test comment carried a correction saying the
framing's wording was wrong. It was wrong only against that setup —
and §9.2's narrowing would then have left the setup with no bite at
all, since a non-recursive delete no longer inspects a descendant.
So the buffer is now bound to the EXACT deleted path: a file is
opened, then replaced on disk by a non-empty directory, and
`remove_dir` fails with ENOTEMPTY deterministically under any uid.
Both of the framing's stated pre-images now bite, so the framing
needed no amendment there. The correction is recorded in §9.1 rather
than only in a test comment, which is where the review asked for it.
WHY THE SHIPPED SUITE PASSED WHILE FINDINGS 1 AND 4 WERE LIVE — two
coverage facts for the next lane. Every delete criterion drove the
PRIMITIVE directly, so nothing in the suite ever built a multi-op plan
and the preflight's plan-time behaviour had no test at all; the only
batch test, `m4_15`, happens to delete a path no earlier op touches.
And every recursive-delete criterion (7, 8, 9) passes `recursive =
true`, while every non-recursive one binds its buffer to the exact
target, so no test in the suite ever combined a non-recursive delete
with a descendant buffer — the exact cell finding 4 lives in.
Sweep, per the review's request. Seven sites decide something is
"affected"; the table is in framing §9.7. Three were the defects
above. Two are unchanged by design and named so they are not mistaken
for oversights: phase-4 reconciliation compares paths RAW via
`BufferRegistry::find_by_path`, which Q#RD10 pins as "exactly today's
behaviour" and which correcting would widen reconciliation — the one
thing Q#RD5 forbids; and `delete_verdict` stats the raw path while
comparing normalized ones, a latent inconsistency whose every branch
fails safe and which matches the primitive's own `remove_file`. Two
are consistent: the `_delete_verdict` binding defaults `recursive` and
`ignore_if_not_exists` the same way the primitive does, and the
deferral set is enumerated (create: 1 path; rename: 2; delete: 1;
edit: excluded, with the argument written down) rather than reasoned
about. Nothing else in the lane decides an affected set.
Bites. Every row was RUN, with the positive control `scripts/bite`
gained in #192 (merged into this lane), and every ref-based row below
reports `OK (assertion)` rather than `OK (COMPILE)`. `1873be6` is this
lane's own first commit: findings 1, 3 and 4 were introduced by it, so
`main` cannot falsify their pins.
rd11a builtin/runtime/lsp.lua @ main OK (assertion)
rd11b builtin/runtime/lsp.lua @ main OK (assertion)
rd11c builtin/runtime/lsp.lua @ main OK (assertion)
rd11d builtin/runtime/lsp.lua @ main OK (assertion)
rd12a builtin/runtime/lsp.lua @ main OK (assertion)
rd12b builtin/runtime/lsp.lua @ main OK (assertion)
rd13 builtin/runtime/lsp.lua @ main OK (assertion)
rd15 builtin/runtime/lsp.lua @ main OK (assertion)
rd18 src/lua_bindings/mod.rs @ 1873be6 OK (assertion)
rd19a builtin/runtime/lsp.lua @ 1873be6 OK (assertion)
rd19b builtin/runtime/lsp.lua @ 1873be6 OK (assertion)
rd19c builtin/runtime/lsp.lua @ 1873be6 OK (assertion)
rd20 builtin/runtime/lsp.lua @ 1873be6 OK (assertion)
Two rows need their weakness stated rather than hidden.
rd11 is VACUOUS against `main`'s `lsp.lua` and the script says so — a
preflight-less applier passes it, which is expected, because rd11 is
the direction that asserts the guard does NOT over-refuse (the same
shape as criteria 2, 7, 9 and 14). It bites two other ways, both run:
`OK (assertion)` against `main`'s `src/lua_bindings/mod.rs`, where the
primitive's absent-plus-ignore branch destroys the buffer; and against
a hand mutation dropping `ignore_if_not_exists` from the preflight
call, which is the pre-image the framing actually names for it.
rd3's two pre-images are designs never committed, so no ref carries
them and `scripts/bite` cannot be used. Hand-mutated instead:
reconciliation moved ahead of the filesystem mutation makes rd3 fail
on exactly its stated assertion (and rd4 with it). On this setup that
mutation and "validation that removes rather than inspects" are the
same mutation, because the buffer is bound to the exact deleted path —
stated because the first shipped setup could see neither.
The eight rows against `main`'s `lsp.lua` all fail by TIMEOUT rather
than by a value assertion, and that is the pre-image behaviour, not a
flaky harness: on `main` the primitive's raise escapes the applier,
escapes `handle_server_requests`, is swallowed by the
`pcall(handle_server_requests)` at the bottom of the file, and the
server is never answered at all. The sink is therefore never written.
That unanswered request is the defect criterion 13 exists to pin.
Gates: fmt; clippy -D warnings; --lib 1863; --lib --features crdt
2048; m4_acceptance 146 (was 132); lsp_dispatch_seams_acceptance 15;
dired_acceptance 25 and autosave_acceptance 29 (the framing's watch
items); PMACS_REQUIRE_GPU=1 -p pmacs-gpu 202; git diff --check clean.
No protocol change.
Implements the framing merged as #186. On `main` today,
`pmacs.buffer.apply_resource_op`'s delete arm removes a file and then
removes any buffer bound to it, with no dirty check at any link in the
chain — so a server-driven delete destroys unsaved edits, and the
`ignore_if_not_exists` arm destroys them having done no filesystem work
at all.
Layer 1 — the primitive. The delete arm becomes four ordered phases:
stat/no-op decision, enumerate and validate, mutate the filesystem,
reconcile the registry. Validation inspects and removes nothing, so a
filesystem failure leaves every buffer intact automatically rather than
by compensation, and `on_removed` still observes the path already gone
because reconciliation stays last.
`delete_verdict` is the single shared query. It scans *every*
path-bound buffer rather than the first match, because `find_by_path`
is first-match-only and `pmacs.buffer.from_file` makes duplicates
reachable — a clean first match could otherwise hide a modified second.
It normalizes both sides before comparing and uses component-aware
`starts_with`, so `/tree` does not match `/tree-sibling`. It stats with
`symlink_metadata`, not `canonicalize`, which reports a dangling
symlink as absent and would disagree with the primitive on exactly the
input `ignore_if_not_exists` turns on.
Layer 2 — the applier and the server-request boundary.
`apply_workspace_edit` gains a plan-time delete precondition check
driven by the same Rust helper, so the two layers cannot drift. It is a
filter, not a transaction, and the code says so: `documentChanges` are
sequential, so an earlier edit can dirty a buffer a later op deletes.
The applier is now total — every failure becomes `nil, message`, and
the origin buffer is restored on the failure path as well as the
success path. At the boundary, parse *and* apply are wrapped:
`_parse_workspace_edit` sits one line above the applier and is
fallible, so a parse failure previously escaped, was swallowed by
`pcall(handle_server_requests)`, and left the server unanswered — the
defect being fixed, one line out of scope. Failures now also append one
labelled record to `*errors*`.
Scope, stated plainly rather than implied by what is present:
* Acceptance criteria 1-10, 14 and 16 land here — 11 tests driving
the primitive directly. Criteria 11, 11a-11d, 12, 13 and 15 do
NOT: they exercise Layer 2 through a real server pump and need
`pmacs_fake_lsp` modes that do not exist yet. Criterion 13
explicitly rejects a direct-call test as insufficient, so the
Layer 2 code currently has no production-path pin. That is a real
gap and the reason this is not the whole lane.
* The framing's §8 branch plan said the implementation would land on
#186 itself. #186 merged as framing-only, so it gets its own
branch and PR. No decision changes.
* Criterion 3's stated bite in the framing is wrong. It claims to
fail against buffer-first ordering; it does not, because the
deleted path is a directory no buffer is bound to, so the
reordering never fires on that input. It does fail against
validation that removes rather than inspects. Checked by mutation
rather than trusted, and the test comment carries the correction.
Bite: criteria 1, 5, 6, 8 and 10 fail against `githubsucks/main` under
`scripts/bite`. Criteria 3 and 4 pin phase ordering against designs
never committed, so `main` cannot falsify them; both were verified by
hand mutation instead. Criteria 2, 7, 9 and 14 assert preserved or
deliberately-unchanged behaviour and pass against `main` by design —
2 is criterion 1's opposite direction, 9 pins today's imperfect
orphaning so widening cannot happen silently.
Gates: fmt; clippy -D warnings; --lib 1863; --lib --features crdt
2048; m4_acceptance 132; lsp_dispatch_seams_acceptance 15;
dired_acceptance 25 and autosave_acceptance 29 (the framing's watch
items); PMACS_REQUIRE_GPU=1 -p pmacs-gpu 202; git diff --check clean.
No protocol change.
Make command-time attachment healing cancel an armed terminal restart
before replacing the server, while keeping request-only lookup pure and
restart-safe.
Track config-driven server ownership privately, bound every fallback
server per SID, scope no-swap retirement to the failed root, and route
the shipped Lean diagnostics command through the safe resolver while
waiting for initialization.
Add direct acceptance counterexamples for all five review findings and
record the sixth-round verification and vacuity lesson.
Round 5 review: one P1, a frontend scope hole, and three P2s.
**1. A fallback that SPAWNS and then dies retried forever.** The
once-per-buffer guard bounds calls to `_attach_buffer`, not the server
those calls produce. `ensure_server` still never forwards `cfg.restart`,
so the fallback inherits `OnCrash`; an executable that exits before
`initialize` is respawned by the manager with no attempt ceiling —
silently, because `latched` has already disabled the primary's failure
poll. The fallback now gets its own one-shot die-before-initialize
watch, which retires it (ending the respawn loop) and reports.
The prior failing-fallback test used a NONEXISTENT executable, so it
only ever exercised synchronous ENOENT. To reach "spawned, then died"
the fixture has to actually spawn.
**2. Simultaneous frontends.** Both repair triggers read the ambient
`pmacs.window.buffer()`, and the daemon restores `active_frontend` to
the last-dispatched frontend before `tick_processes` — so a Lean buffer
active in ANOTHER frontend receives no `buffer.after-switch` here and
stays stale after its server is globally retired.
Fixed at the seam that is frontend-agnostic: **make consumption safe.**
`attached_for_active` now rebuilds rather than returning a record whose
server is dead, and `attachment_for_request` reports none (it must not
perturb LSP state, so it cannot rebuild). Whichever frontend runs a
command is the active one while it runs, so healing at the point of use
reaches every buffer no eager sweep can. This also closes the half where
a dead attachment was handed to a command and the request vanished.
**3. The retirement sweep stopped user-managed servers.** Selecting on
`language_id == "lean4"` also names servers the user spawned from
`init.lua`, which are not derived from `pmacs.lsp.config.lean4`. It now
keys on the `default-lean4` label `ensure_server` stamps — the
derivation discriminator.
**4. Repair ran even when no swap occurred.** `swap_to_fallback()`
returning false left `latched` true, so the next tick retried the
UNCHANGED configuration and reported it as a fallback failure. Split
into `probe.fallback_installed`: repair exists to apply a swap, so no
swap means nothing to apply.
**5. The once-per-buffer assertion counted table keys**, which cannot
distinguish "once per buffer" from "every tick for one buffer" —
cardinality stays 1 either way. Replaced with a numeric attempt counter;
the bite reports 174 attempts against the expected 1.
Five bites, each against 7c37bdc: no fallback watch -> attempt reaches
4; retire by language_id -> the user's server is stopped; gate repair on
`latched` -> a repair is attempted with no swap; drop the
once-per-buffer guard -> 174 vs 1; hand back a dead attachment -> a
command receives a `stopped` server.
Two more vacuity shapes recorded in the ledger (8 and 9): counting
distinct keys cannot bound repeated work, and a nonexistent executable
cannot reach any post-spawn failure.
Round 1 review, four P1s. All real; the first two mean the fallback did
not work at all.
**1. The latch swapped the config but never spawned or re-attached.**
Nothing re-fires an attach on a config change and `attach_buffer`
early-returns for a live attachment, so the buffer stayed bound to the
server that had just been stopped. The user got a config edit and no
language server. `fire_latch` now rebuilds through a new
`pmacs.lsp._attach_buffer` export.
Two mechanics had to be right for that rebuild to happen at all:
* It is **retried on the tick**, because `pmacs.lsp.stop` leaves the
state `shutting-down`, which `server_is_live` counts as LIVE — an
inline re-attach early-returns the stale record and the swap is a
silent no-op.
* The latch **does not stop an already-terminal server**, and this is
a substrate bug worked around rather than a style choice.
`LspManager::stop` on a `Crashed` client takes its not-initialized
branch, terminates the dead process, and sets `ShuttingDown { ..
None }` on the premise that "the next exit observation cleans up" —
but the exit already happened, which is what made it `Crashed`. No
further event arrives, so the client is stuck in `ShuttingDown`
forever: `server_is_live` reads it as live so `attach_buffer` never
rebuilds, and `forget` refuses it for not being terminal. Stopping a
dead server is what makes it un-replaceable. Named in framing §6; the
fix belongs in `stop` and changes behavior for every language.
**2. A missing `lake` bypassed probe and latch entirely** — the single
most likely real failure. `ensure_server` swallows a synchronous ENOENT
and returns nil, so there was no attachment, and the hook keyed on
`active_attachment()` returned before arming anything. The hook now keys
on the buffer's LANGUAGE and treats a Lean buffer with no attachment as
the failure itself.
**3. `waitForDiagnostics` omitted `version`.** Lean's
`WaitForDiagnosticsParams` is `{ uri, version }` (v4.9.0,
`src/Lean/Data/Lsp/Extra.lean`); the request is how a client says which
revision it wants. It looked correct only because the fake server echoes
any payload — so the fake server now validates and returns InvalidParams
without it.
**4. The ledger stated the dangerous stacking order** in one sentence
and the correct rule in the next. Fixed to say BEFORE. A safety rule
written twice with opposite senses is worse than not written.
Also (P2): the probe/latch suite now drives the production path —
`buffer.after-load` -> ticks -> probe drain -> latch -> re-attach — with
real executable stubs, and asserts the originally opened buffer ends up
on a LIVE server. Round 1's acceptance 36 asserted every server was
terminal, i.e. pinned the ABSENCE of the fallback it claimed to test.
`M.fallback` is a table so the suite can point it at a working stand-in;
the probe now spawns `cfg.command --version` rather than a hardcoded
`lake`, which is also more correct for a user who configured a wrapper.
`swap_to_fallback`'s `command ~= "lake"` guard is gone: the latch fires
only when the configured server actually failed, one visible fallback
beats no server, and `probe.latched` is what keeps it to exactly one.
Three new bites, all against the committed tree: no re-attach after the
swap -> three latch tests fail; hook keyed on the attachment -> the
missing-`lake` case fails; `waitForDiagnostics` without `version` ->
acc37 fails with the server's InvalidParams.
Six findings from review, one of them a real defect.
**`canonicalize` could emit a path that exists nowhere.**
`p.display().to_string()` substitutes U+FFFD for non-UTF-8 bytes, so a
resolution landing on such a path returned a plausible-looking string
that does not exist on disk — worse than nil, because this value becomes
a server-affinity key through `file_uri_for` and would silently fail to
round-trip, while the doc promised nil for anything unresolvable. Now
`.and_then(|p| p.to_str().map(str::to_owned))`: unrepresentable is a
decline, matching how the fs layer already treats non-UTF-8 symlink
targets.
Pinned by a new acceptance case that reaches a non-UTF-8 target through
an **ASCII** symlink, so the input is representable and only the
resolved output is not — the case a UTF-8 check on the argument would
miss. Bitten: restoring `display()` fails it.
The other five:
- `on_response`'s doc comment now warns that registering against a
server with no attached buffer is fire-on-death, not fire-on-reply,
because the drain visits only attached sids. It looks exactly like a
hung request while debugging, and 3b is the first caller likely to
hit it.
- `deliver_response`'s comment still carried the pre-correction
rationale ("removed BEFORE invocation ... must not be re-entered") —
the claim the bite disproved. It now says what is true: removal is
unconditional, before-vs-after is unobservable without a re-entrant
drain, and the reachable bug is gating removal on a clean return.
- Dropped `server_attempt`'s unused second return.
- Deleted a vacuous assertion in the no-attachment test (counting `_G`
entries to assert "lua globals are readable") — scaffolding that
pinned nothing, the exact shape the project's own lesson flags.
- `#[cfg(unix)]` on the three symlink-dependent tests.
Gates re-run in full. The sweep's first pass tripped
`composition_overhead_under_ten_percent` at 18.8% against a 10% budget;
it passes 3/3 in isolation here, passes in isolation on main, and the
same run reported the realistic-frame overhead as **-4.6%** — a negative
figure is measurement noise, not added work. Nothing in this diff is on
the render path. Rerun of the full sweep: 3,189 across 93 suites, zero
failures.
Two corrections found by bite-testing the suite rather than by reading
it.
**Acceptance 32 was mis-named.** It claimed to pin "the one-shot is
removed BEFORE invocation". Biting that — moving the removal after the
`pcall` — still passes, because `pcall` catches the raise either way and
the removal runs regardless. The before/after ordering is unobservable
unless a handler re-enters the drain, and nothing does. What the test
actually pins is that removal is **unconditional**: the bite that gates
it on `if ok then` fails 2 != 1, because the surviving registration is
invoked a second time by the purge. Renamed and re-commented to say so.
The implementation still removes before invoking, which is the right
defensive order; it is simply not what the assertion proves.
**The purge's generation check is defensive and untested**, now labelled
in place instead of reading as covered. Reaching it needs a crash and
its restart to both land in a gap with no `_async.tick`; the backoff is
500ms, so any tick in that window sees `crashed` and the
absent-or-terminal test fires first. Every attempt to stage it
deterministically exercised the `crashed` path instead. It stays as
insurance for a stalled editor, and says that about itself.
Bites recorded, all against the committed tree:
- removal gated on a clean return -> acc32 fails (2 != 1).
- purge driven by a `crashed`/`stopped` event seen in the drain, the
design the framing originally implied -> the no-attachment case fails
("never called"), while the attached case still passes. That is the
discrimination acceptance 34's second half exists for.
- a resolver without `pmacs.fs.canonicalize` -> two servers, pinned as
34b's own falsification.
Arc 8 Stage 3a (framing Q#LN9, Q#LN20). No Lean content: this changes
the event drain every LSP language runs through, and is split from the
Lean server work for the reason Stage 2 was.
**The seams.** `handle_server_requests` handled five `request` methods
and `initialized`, dropping every `notification` and `response` on the
floor. Dropping responses made `pmacs.lsp.send_request` a write-only API
from Lua — the reply was drained and discarded, so nothing outside
Rust's typed stores could consume one. Two new arms route to
`pmacs.lsp.on_notification(method, fn)` (persistent, method-keyed) and
`pmacs.lsp.on_response(sid, request_id, fn)` (one-shot). Both extend the
existing loop rather than opening a second `events_take` caller, which
would steal events from it.
A one-shot is removed **before** invocation, so a raising handler cannot
be re-entered. Every subscriber is `pcall`ed and a raise reports through
`pmacs.editor.set_status` per COHERENCE §1.2 — not `pmacs.error`, which
is defined nowhere in production. The notification list's length is
captured before the walk so a subscriber registering another cannot
extend the list being iterated.
**The purge is driven off `pmacs.lsp.list()`, not off a death event.**
The framing said acceptance 34's second edge was a killed buffer. That
was wrong, and scouting the implementation is what caught it: pmacs
fires exactly five hooks (`buffer.after-edit`, `buffer.after-load`,
`buffer.after-switch`, `frontend.detached`, `process.after-tick`) and
there is no buffer-kill hook at all, so `lsp.lua` never tears an
attachment down and the drain keeps reaching that server. No leak there.
The real leak is a different path with the same root cause. The drain
builds its sid list from `attachments`, and `attach_buffer` drops a sid
from that table the moment `server_is_live` reports false — rebuilding
against a fresh server. So the `crashed` / `stopped` event that should
trigger the purge is precisely the one most likely to go undrained. A
purge wired to that event leaks exactly when it matters.
`pmacs.lsp.list()` enumerates the manager directly and is unaffected by
attachment bookkeeping, so the purge polls it after each drain: a sid
that is absent, terminal, or running a **new generation** settles its
pending one-shots with an error. The generation check uses the `attempt`
field, because a crash-then-restart reuses the sid — without it a
one-shot would sit waiting on a reply the dead generation owed.
**`pmacs.fs.canonicalize`** (Q#LN20) is the one synchronous function on
`pmacs.fs`, and synchronous is the point: its consumer is a
function-valued `config.root` called from `ensure_server` <-
`attach_buffer` <- `buffer.after-load`, where there is no coroutine and
`pmacs.fs.stat`'s awaitable handle is unusable. It is installed from
`install_async` rather than `install_project` purely for load order —
`make_workspace` runs after `fs.lua` is evaluated, so a canonicalizer
placed there reads nil.
Acceptance: `tests/lsp_dispatch_seams_acceptance.rs`, 14 tests, driven
against `pmacs_fake_lsp` through rust so nothing needs a toolchain.
Dispatch integrity is exercised at real co-occurrence — the fake server
writes `workspace/applyEdit` and the `executeCommand` reply back to
back, so both land in one `events_take` batch. 34b asserts affinity
survives a symlinked open and is paired with its own falsification: the
same resolver minus the canonicalize call spawns two servers, so the
positive case cannot be vacuous.
COHERENCE.md §1.2 makes "a `pcall` around background wiring must log
attributed failure, never discard it" a standing rule, and names
`ensure_server`'s swallowed spawn failure as its canonical case — the
exact function this branch modifies. Round 1 deferred the resolver's
silent `pcall` as a Stage 3 concern. Under that rule it is not a
deferral, it is a fresh instance of the named anti-pattern added by a PR
touching the cited function, made worse by the memo: a raised error is
buried permanently for that directory and never observed again.
A resolver that raises, or returns a non-string non-nil, now leaves an
attributed trace naming the language and the directory. Returning nil
remains the documented decline and stays silent — pinned, so "report
failures" cannot be satisfied by reporting every resolution.
The report goes through `pmacs.editor.set_status`, NOT `pmacs.error`,
and that choice is the finding:
**`pmacs.error` does not exist.** Fifteen call sites across `async.lua`
(5), `syntax.lua` (4), `lsp.lua`, `mcp.lua`, `fs.lua`, `editops.lua`,
`autosave.lua`, and `commands/default.lua` report background failures
through it, each guarded `if pmacs.error then ...`. It is defined
nowhere in production; the only assignment in the tree is a test stub at
`src/editor.rs:9881`, and `type(pmacs.error)` is nil in a fresh
`EditorState` (probed, not inferred). `pmacs.errors` (plural) in
compile.lua is an unrelated namespace. So all fifteen reports are dead,
and the guard makes the silence look deliberate — which is why nobody
noticed. Writing the test is what caught it: the first version of this
fix used `pmacs.error` and its pin failed against a working
implementation.
Both bites recorded: dropping the report entirely fails the pin, and so
does reporting ONLY through `pmacs.error` — the dead-channel variant
this nearly shipped.
Not fixed here, deliberately: defining `pmacs.error`, the fifteen dead
sites, and surfacing the spawn failure itself. That last is Priority 1
work and a user-visible product behavior — what message, where, with
what guidance — so it needs its own framing rather than being smuggled
into an affinity PR.
The blocker was process, not design. The test file was committed before
`cargo fmt` ran, so the reflow of five over-width assertions sat
uncommitted in the working tree while the branch as pushed failed the
first gate in CLAUDE.md. The "fmt clean" reported on the PR described
the worktree, not the branch. Gate results are only meaningful run
against the pushed tree, so this commit lands the formatting first and
the gates are re-run against it.
Two pins review asked for, each covering a branch the nine acceptance
tests left untested:
- A **string** `config.root` as an affinity key. acc17 covers only the
function form, so `return configured, "config"` had no test. The bite
puts both files in their own marked project: drop the config arm and
they key on their own detected roots and spawn two servers, so one
server on the configured root is only reachable if the override wins.
- `root = false` reads as unset. Defended by a truthiness check rather
than `~= nil`, previously by comment alone. Under `~= nil` the config
arm returns `false, "config"` and `file_uri_for(false)` returns nil, so
the file lands on a rootless server instead of its detected project.
Each was falsified against exactly the mutation it targets and neither
against the other.
Also documents an asymmetry review caught: `project_root_for`'s
"detected" arm is canonicalized for free because `pmacs.project.detect`
canonicalizes before walking, but a **configured** root — string or
resolver return — is fed to `file_uri_for` exactly as written, and the
affinity key is that URI. On macOS a resolver returning `/var/…` and a
detected `/private/var/…` are therefore different keys for one
directory, silently yielding two servers for one project. There is no
Lua-side canonicalizer to normalize it, and Stage 3's Lean resolver is
the first real consumer, so the obligation is stated in the
`config.root` doc comment where that resolver's author will read it.
`ensure_server` reused any live server whose `language_id` matched,
regardless of project root — its own comment documented this as a known
post-v0.1 limitation. For project-model-strict servers that is a
correctness failure, not a rough edge: `lake serve` is bound to one Lake
package, rust-analyzer and gopls to one workspace, so the second project
a user opens gets a server that cannot resolve its imports.
Server affinity is now keyed on the project root, with one rule that
keeps the change from regressing every other language:
The affinity key is the root only when a root was actually FOUND.
`project_root_for` never returns nil for a file that has a path — its
last resort is the file's own directory — so a naive `(language_id,
root)` key would give every directory of loose scratch files its own
server, for every language: two stray .py files in different directories
would spawn two pyrights where today they share one. It now returns
`root, source` with source one of "config" / "detected" / "fallback",
and only the first two become an affinity key.
Matching is on the spawned spec's `root_uri`, nil matching nil, so the
fallback spawn passes `root_uri = nil` for the key and the stored spec to
agree. `cwd` still carries the directory, and `build_initialize`
(src/lsp.rs) derives the identical `rootUri` from `cwd` when the field is
None — using a percent-encoder with the same allowed set as Lua's
`file_uri_for`. The initialize payload for that case is therefore
byte-identical to before; only what the reuse loop matches on changes.
`build_initialize` is the only reader of `spec.root_uri` in the tree.
Two consequences, both deliberate and both asserted rather than
discovered:
- A server hand-spawned from init.lua with only `cwd` set also reads
back nil, so a root-bearing attach will not adopt it. We cannot know
which root it was meant to serve, and guessing wrongly routes a
project's files to the wrong server.
- Opening files across N project roots spawns N servers. rust-analyzer
has the same property and no editor caps it by default; `pmacs.lsp.stop`
is the manual escape and an LRU reaping policy stays deferred.
`config[language].root` may now be a `function(path) -> string|nil` as
well as a string, for languages whose root rule the shared marker walk
cannot express — an innermost-wins walk cannot find an *outermost*
marker. A resolver returning nil declines and falls through to the marker
walk. Results are memoized per directory because hoisting the root
computation above the reuse loop puts it on every attach rather than
every spawn; the memo is keyed weakly by the resolver function itself, so
replacing `config[lang].root` cannot serve a root the old one computed.
`pmacs.lsp.list()` rows gain `root_uri` and `cwd`. `root_uri` is the spec
field verbatim, deliberately not the URI the server was initialized with.
No protocol change. No Lean content: this is the shared affinity function
for every LSP language, so it ships as its own PR and is exercised
through rust, python, go and typescript against `pmacs_fake_lsp`.
tests/lsp_multi_root_acceptance.rs covers acceptance 13-21. Every fixture
sets `pmacs.project.set_search_boundary` at its own tempdir root:
without it the marker walk climbs to the filesystem root, and a stray
`.git` above the temp directory would turn the markerless cases into
detected ones — the assertions would still pass while testing nothing.
Refs docs/lean4-mode-framing.md Q#LN15, acceptance 13-21.
- `listview.open`, `compile.run`, and `pmacs.terminal.open` all take the
same strict `display = "current" | "panel"`, validated before any
buffer, session, process, or wrapper exists. Omission keeps today's
behavior; Stage 3 flips the default.
- `listview.quit` / `compile.quit` delegate to `window.quit` only when
the buffer really is in a side window, so the presentation is deleted
or restored instead of leaving a source buffer stranded in the slot.
- LSP `visit_location`, LSP go-to-definition, and compile `visit_error`
route through `display_file`, so a visit from a panel lands in the
document target and fires its hook with that window active.
- `window.quit`'s Delete arm focuses the revalidated remembered origin.
- Capability fallback discards an accompanying `height` rather than
rejecting the call.
- `window.min-height` clamps a below-floor value on read instead of
refusing the write.
- `tests/bottom_panel_stage1_acceptance.rs`: 42 tests over the framing's
Stage 1 criteria, including the two production `Layout::compute`
callers, the recursive minima, hide/reappear, the final-focus matrix,
quit chains at the depth cap, per-frontend jump origins, the divider,
and a real-PTY pin of Bet B1.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Parse bounded Emacs and Vim modelines, normalize common aliases, and give
explicit file metadata precedence over inferred language. Pin one fresh-load
language decision for syntax, LSP, pairing, comments, and initial major mode,
while preserving the LSP path guard and explicit mode overrides.
Cover supported forms, rejection boundaries, precedence, unknown modes,
shebang and modeline pinning, reopen behavior, and pathless buffers.
Add the strict pmacs.statusline provider registry, deterministic
borrow-released per-window evaluation, context-scoped failure latches,
and a pure built-in LSP provider.
Preserve the legacy TUI modeline while composing faced custom runs,
and append authoritative complete StatuslineSegments replacements for
semantic frontends. Expand dynamic ThemeFacts, reset producer/frontend
baselines symmetrically, and gate all provider work off protocol v18.
Teach the GPU to atomically validate, resolve, shape, clip, and cache
custom modeline runs without displacing the protected status suffix.
Document the public Lua lifecycle, wire ownership, snapshot semantics,
and the fully gated Arc 4 stage-3 delivery state.
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Drive Red Hat yaml-language-server 1.24.0 through the default YAML
auto-attach path. Disable SchemaStore and the Kubernetes CRD catalog for
network-free determinism, require language-specific initialization and a
real syntax diagnostic, and prove the server remains alive afterward.
Update the framing and runtime commentary with the completed live-provider
evidence. The test passes against the pinned provider and fails against the
pre-JSON/YAML runtime under scripts/bite.
Co-Authored-By: OpenAI Codex <noreply@openai.com>
Preserve PR #123's unpushed review fixes on a transfer branch: initial
didChangeConfiguration delivery, explicit JSON validation, the pinned
JSON server provider, corrected YAML configuration sections, and
deterministic plus real-provider acceptance coverage. Record the
observed yaml-language-server 1.24.0 standalone smoke and leave the
real YAML-through-pmacs test, rebase, and full gates explicitly pending
for the destination machine.
Add tree-sitter-json (0.24) and tree-sitter-yaml (0.7) to
BUILTIN_LANGUAGES (both ABI-current via tree-sitter-language, verified
compiling under tree-sitter 0.26), each self-contained highlights, no
injections of their own. Extensions json=.json, yaml=.yaml/.yml; root
kinds json `document`, yaml `stream`.
The payoff from the #122 injection engine is free: the markdown block
injection query already sets injection.language "yaml" for `---`
frontmatter (minus_metadata) and "toml" for `+++` (plus_metadata), so
registering yaml lights up YAML frontmatter highlighting with no extra
wiring, and ```json / ```yaml / ```yml fences resolve through the engine
(yml->yaml alias already present). Two acceptance tests pin this synergy.
LSP (builtin/runtime/lsp.lua): pmacs.lsp.config.json uses the maintained
extracted-bundle binary `vscode-json-language-server --stdio` (NOT the
stale standalone vscode-json-languageserver); MIT, no telemetry, remote
$schema fetch left enabled (no handledSchemaProtocols). pmacs.lsp.config
.yaml uses `yaml-language-server --stdio` with Red Hat telemetry
disabled by default. Both ship the exact workspace/configuration sections
each server pulls (json+http; yaml+http+redhat.telemetry) present-not-null
so the servers get defaults rather than erroring — the CMake #117 lesson.
Sections derived from server source/docs (neither binary installed on
this build machine to observe live; verify where present). Filetype
fallback entries added. JSON is the standing prerequisite for the Jupyter
.ipynb arc; handoff §6 updated.
Nine acceptance tests (grammar ABI, highlights compile, detection,
grammar<->LSP-key alignment, the two frontmatter/fence synergy proofs,
and the pinned LSP-config sections).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJ9FQ832QwftJXCD9LeFan
[P2] The shared JavaScript highlights query guards its builtin captures
(console, require, …) with `#is-not? local`, a PROPERTY predicate
(`Query::property_predicates`) that needs a scope map from the grammar's
LOCALS_QUERY — which pmacs does not run. `compute_highlight_spans` took
every capture, so a locally-shadowed `console`/`require` still surfaced
as `@variable.builtin`/`@function.builtin`; a theme distinguishing
`.builtin` would mis-style the shadowed local.
Full locals processing is substrate work; conservatively fail-closed
instead: drop captures whose pattern carries an `#is?`/`#is-not? local`
property predicate (the identifier falls back to its non-builtin
capture). The text predicates (`#eq?`/`#match?`/`#any-of?`, already
applied by the capture iterator) and `#set!` settings are untouched.
This is a general engine fix — it corrects the same latent mis-styling
for any grammar using the locals predicate, not just JS/TS.
- javascript_shadowed_builtin_is_not_mislabeled: a local `const console`
produces no `*.builtin` capture (directly observed to fail — two
`variable.builtin` captures — before the fix).
[P3] Comments this PR invalidated: `lsp.lua` no longer claims Python has
no grammar; `syntax.lua`'s `_has_language` gate comment uses a
still-grammarless example (an init.lua `shebangs.ruby`) instead of
python/javascript; and the rewritten Ruby shebang test's doc no longer
describes it as a Python test.
Gates: fmt; clippy -D warnings; test --lib; --features crdt;
m4_acceptance --skip basedpyright; GPU; full workspace sweep;
git diff --check.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJ9FQ832QwftJXCD9LeFan
cmake-language-server does NOT pull a `workspace/configuration` section:
it reads `buildDirectory` from the `initialize` request's
`initializationOptions`, and drives its project model off CMake's File
API under `<buildDirectory>/.cmake/api/` (not `compile_commands.json`).
The `settings = { cmake = {} }` block — and the documented
`settings.cmake.buildDirectory` override — were therefore inert, leaving
conventional out-of-source project data unavailable.
Replace it with `init_options = { buildDirectory = "build" }` (the
conventional out-of-source dir; users override `init_options.buildDirectory`
from init.lua), and correct the comment. The wiring test now asserts
`config.cmake.init_options.buildDirectory == "build"` — bite-verified
against the pre-fix lsp.lua.
Gates: fmt; clippy -D warnings; --features crdt (1720); m4_acceptance
--skip basedpyright (109); GPU (59); full workspace sweep (zero
failures); git diff --check — all green. One `--lib` run flaked on
process::m6_1_pty_mode_lifecycle_started_then_exited (PTY-lifecycle
timing, the m6/m8 daemon-timing family); it passed in the crdt run, the
full sweep, and 4/4 isolated — unrelated to this Lua config change.
Change is Lua config + the acceptance assertion.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJ9FQ832QwftJXCD9LeFan
Files identified by their whole basename — Dockerfile, Makefile,
CMakeLists.txt, rc dotfiles — had no detection path (extension-, then
shebang-keyed). Add a filename layer and the three grammars behind it.
- **Grammars** (BUILTIN_LANGUAGES): dockerfile via `tree-sitter-containerfile`
(the ABI-current grammar; the old `tree-sitter-dockerfile` pins
`tree-sitter ^0.20` and would fork the graph — containerfile rides
`tree-sitter-language 0.1`, tree-sitter dev-only, like the others),
make via `tree-sitter-make`, cmake via `tree-sitter-cmake`. All ship
self-contained highlights (single fragment). Extensions:
`.dockerfile`/`.containerfile`, `.mk`/`.make`, `.cmake`.
- **Filename layer**: `pmacs.parse.language_from_filename(name)` backed by
an extensible `pmacs.parse.filenames` map, wired into the precedence
chain in both syntax.lua (grammar) and lsp.lua (LSP): grammar-ext →
filetype map → filename → shebang. A recognized extension still wins;
the basename map only fires when the extension misses. Seeds the three
filenames plus shell rc dotfiles (`.bashrc`/`.zshrc`/`PKGBUILD`/… →
bash) — highlighting them against the grammar shipped in #115.
- **LSP**: `config.dockerfile` (docker-langserver --stdio) and
`config.cmake` (cmake-language-server). Make has no server, so no
`config.make` — grammar highlight only. Extension filetype fallbacks
added for id stability.
Bite-verified acceptance:
- filename_grammars_load_and_parse — each grammar's ABI is accepted by
the tree-sitter 0.26 core and parses a representative snippet without
error (dockerfile/cmake root at source_file, make at makefile).
- builtin_languages_include_dockerfile_make_cmake /
language_for_path_resolves_dockerfile_make_cmake_extensions — entry
presence and extension detection.
- m4_filename_map_resolves_special_files — the basename map (incl. path
form and dotfiles→bash), config.dockerfile/cmake commands, and no
config.make. Bite-verified against pre-feature syntax.lua.
- m4_filename_extensionless_dockerfile_highlights — an extensionless
`Dockerfile` resolves to dockerfile for LSP and gets a dockerfile parse
tree; reachable only via the filename map. Bite-verified.
Gates: fmt; clippy -D warnings; test --lib; --features crdt;
m4_acceptance --skip basedpyright; GPU; full workspace sweep;
git diff --check.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJ9FQ832QwftJXCD9LeFan
Extension detection missed extensionless scripts — `scripts/deploy`, git
hooks, `configure`, and `scripts/bite` itself — so they got neither
highlighting nor an LSP server. Add a first-line shebang fallback.
- New `pmacs.parse.language_from_shebang(buf)` (builtin/runtime/syntax.lua):
sniffs the first line (capped at 256 bytes), maps the interpreter's
basename to a language, and resolves the `#!/usr/bin/env python3`
indirection (skipping env's own `-S`/flags and `VAR=val` assignments).
Backed by `pmacs.parse.shebangs`, a user-extensible map seeded with the
interpreters pmacs can act on: sh-family -> bash, python* -> python,
node -> javascript, lua* -> lua.
- Wired as a strict *fallback* on both resolution paths: syntax.lua's
grammar attach (`language_for_path or language_from_shebang`) and
lsp.lua's `buffer_language` (grammar -> filetypes -> shebang). A
recognized extension always wins, so a `.py`/`.sh` file is never
re-classified by a stray shebang.
- Cross-language, not shell-only: `#!/usr/bin/env python` /`node` /`lua`
resolve too. Special filenames (`.bashrc`, `Dockerfile`, `Makefile`)
are intentionally deferred until there are grammars behind them.
Bite-verified acceptance (tests/m4_acceptance.rs):
- m4_shebang_resolver_maps_interpreters — the mapping incl. env
indirection and `env -S`; non-shebangs and unmapped interpreters
(ruby) resolve to nil.
- m4_shebang_extensionless_script_resolves_bash — opening an
extensionless `#!/bin/sh` script resolves to bash on BOTH paths:
lsp.lua's `active_buffer_language()` and a settled bash parse tree
(grammar attach). Reachable only via the shebang, since the file has
no extension.
- m4_shebang_does_not_override_extension — a `.py` file opening with
`#!/bin/sh` still resolves to python (extension precedence).
Gates green: fmt; clippy -D warnings; test --lib; --features crdt;
m4_acceptance --skip basedpyright; GPU; full workspace sweep;
git diff --check. Change is Lua-only plus the acceptance tests.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJ9FQ832QwftJXCD9LeFan
Shell scripts already had LSP (bash-language-server + shellcheck/shfmt,
wired in builtin/runtime/lsp.lua), but no tree-sitter grammar, so their
text rendered without lexical color. Fill in the missing half.
- Bundle tree-sitter-bash (0.25) as a BUILTIN_LANGUAGES entry. Unlike
cuda, bash's highlights.scm is self-contained (no `; inherits:`
delta), so a single fragment suffices. The crate exports
LANGUAGE/HIGHLIGHT_QUERY over tree-sitter-language 0.1 — shared ABI
crate, no second tree-sitter in the graph.
- Extension set is wider than the `.sh`/`.bash` the LSP filetype map
covered: `.zsh`/`.ksh`/`.ash` are close-enough dialects and `.bats`
is bash. The grammar's language name is `bash`, matching the
`pmacs.lsp.config.bash` key, so opening any of these also auto-attaches
bash-language-server (shellcheck declines zsh, so `.zsh` diagnostics
may be sparse; highlighting is unaffected). lsp.lua's filetype map is
extended to the same set as the belt-and-suspenders fallback.
- Extensionless shebang scripts (`#!/bin/sh`) and rc dotfiles
(`.bashrc`) are intentionally NOT covered: detection is extension-keyed
and shebang/filename sniffing is a separate, deferred feature.
Bite-verified acceptance:
- bash_grammar_loads_and_parses_script — the 0.25 grammar's ABI is
accepted by the 0.26 core (set_language succeeds at runtime) and a
representative script (shebang, set, parameter expansion, function,
if) parses without error, rooting at `program`.
- builtin_languages_include_bash / language_for_path_resolves_bash_
extensions — entry presence and detection across the wider set.
- bash_highlights_compile_with_captures — the self-contained query
compiles against the grammar with real capture classes.
- m4_12_default_bundle_wires_bash — through the loaded runtime,
config.bash targets bash-language-server and both grammar detection
and the filetype fallback resolve the new extensions to `bash`.
Gates green: fmt; clippy -D warnings; test --lib; --features crdt;
m4_acceptance --skip basedpyright; GPU; full workspace sweep;
git diff --check.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJ9FQ832QwftJXCD9LeFan
Two functional gaps from review:
1. Standalone .cuh files got no clangd AST. clangd selects the
compiler language from the file extension, not the LSP languageId:
it knows .cu (-> -x cuda) but not .cuh, so a header with no compile
command fails with fe_expected_compiler_job. config.cuda now sets
init_options.fallbackFlags = { "-xcuda" }, which supplies -x cuda
for any file this server opens that lacks a compile_commands.json
entry (a real compile command still wins). This CUDA server only
ever serves .cu/.cuh, so the fallback cannot mis-flag C/C++.
2. The CUDA highlights query was only a delta. tree-sitter-cuda's
HIGHLIGHTS_QUERY opens with `; inherits: cpp` and defines only the
CUDA-specific captures (launch brackets, __global__/__device__) —
two capture classes. pmacs does not resolve `inherits:`, so ordinary
C/C++ syntax went unhighlighted. LanguageEntry.highlights_query is
now &[&str] (fragments joined base-first); the cuda entry carries
[c, cpp, cuda], compiling to ~16 capture classes. Fragments are
newline-joined, never bare-concatenated — a fragment can end mid
`; comment`, and abutting the next fragment's first token would
corrupt the query. Existing single-query grammars become one-element
slices (byte-identical effective query; no behavior change).
Tests:
- cuda_highlights_resolve_c_and_cpp_captures — asserts the COMPILED
cuda query carries the C base `@variable` capture and >= 8 capture
classes, not merely a non-empty query (the CUDA delta alone has 2 and
no `variable`, so this fails without the base prepend).
- builtin_languages_include_cuda — now asserts the entry composes the
c + cpp + cuda fragments.
- m4_12_default_bundle_wires_cuda — now asserts
config.cuda.init_options.fallbackFlags[1] == "-xcuda".
Gates green: fmt; clippy -D warnings; test --lib (1515); --features crdt
(1689); m4_acceptance --skip basedpyright (101); GPU (59); full
workspace sweep; git diff --check.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJ9FQ832QwftJXCD9LeFan
Opening a .cu/.cuh file previously resolved to no language, so no
server attached and there was no highlighting. Wire CUDA end to end,
mirroring the existing C/C++ path:
- Bundle tree-sitter-cuda (0.21) as a new BUILTIN_LANGUAGES entry
claiming .cu/.cuh, with its own HIGHLIGHTS_QUERY. A dedicated grammar
rather than reusing cpp: the C++ grammar errors on the
<<<grid, block>>> kernel-launch syntax. The crate rides
tree-sitter-language 0.1 (its tree-sitter dep is dev-only), so it
shares the ABI crate with the other grammars — no second tree-sitter
in the graph.
- pmacs.lsp.config.cuda targets clangd (the same binary that serves
C/C++; language_id "cuda" so clangd enters its CUDA parse mode), and
.cu/.cuh filetype fallbacks map to "cuda" to keep the LSP id stable
if the grammar is ever dropped. LspStyleView layers clangd's CUDA
semantic tokens on top, exactly as for C/C++.
Bite-verified acceptance:
- cuda_grammar_loads_and_parses_kernel_launch — proves the 0.21
grammar's ABI is accepted by the 0.26 core (set_language succeeds at
runtime, which the compile step cannot confirm) and that the entry
wired the CUDA grammar, not a cpp fallback: the <<<...>>> launch
parses without error, whereas the cpp grammar reports an error on the
same source (verified out of band).
- builtin_languages_include_cuda / language_for_path_resolves_cuda_
extensions — entry presence and .cu/.cuh detection.
- m4_12_default_bundle_wires_cuda — config.cuda targets clangd and the
filetype + grammar detection resolve to "cuda" through the loaded
runtime.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YJ9FQ832QwftJXCD9LeFan
Finding 1 (medium): pair entries validate full UTF-8 well-formedness
(Unicode Table 3-7), not just lead-byte length — continuation-byte
shape on every trailing byte, overlong encodings (C0/C1, E0 80-9F,
F0 80-8F), UTF-16 surrogates (ED A0-BF), and beyond-U+10FFFF (F5+,
F4 90+) all disqualify, so "(\xC2x" can no longer inject invalid
bytes as a closer. char_at shares the validator and returns the raw
byte for malformed buffer content: the predicate treats junk as
word-like (no pairing before it), never as EOL. Bite:
malformed_utf8_pair_entries_are_rejected (four ill-formed shapes).
Finding 2 (low): relevance and reporting resolve against the SOURCE
buffer the record names, not whatever buffer a context-switching
command left active. New pmacs.lsp.buffer_language(buf) is the
parameterized primitive (active_buffer_language delegates), backed by
a new buf:path() query on buffer handles. Bites: rust→python `'` now
stays silent; python→rust `'` now reports "source context changed".
Finding 3 (low): non-table set containers degrade
language→default→empty instead of throwing from the after-edit
callback on every keystroke. Bites: a string default pairs nothing
with a clean *errors* buffer; a junk language entry falls back to the
default set.
Framing synced to revision 5 (Q#AP2 well-formedness + container
degradation + source-buffer resolution, Q#AP3 predicate junk-byte
posture, acceptance list).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VoiEyuPjoBhvwACf8HAnLB
New builtin/runtime/comment.lua: `edit.toggle-comment` comments or
uncomments the current line — or every line the region touches — using
the language's line prefix from the public, user-extensible
`pmacs.comment.strings` table (Q#CT3; block comments deferred).
Language detection reuses lsp.lua's grammar+filetypes chain, now
exported as `pmacs.lsp.active_buffer_language()` (the only lsp.lua
touch — one assignment).
Semantics (Q#CT4): uncomment iff every non-blank line already starts
(after its indentation) with the prefix, stripping the prefix plus one
padding space; otherwise comment, inserting `prefix .. " "` at the
minimum indentation of the span's non-blank lines (Emacs comment-region
alignment). Blank lines are skipped in both directions and don't feed
the min-indent; an all-blank span is a status no-op. Mixed spans
comment — the double prefix round-trips, preserving inner
commented-out code.
The whole toggle is ONE buf:replace (Q#CT5): one undo step (no undo
grouping exists — N per-line edits would need N undos), one CRDT op,
and one effective-edit verification with the killring intercept
discipline (pcall'd; a rejection reports rather than throws; any
post-intercept deviation reports and skips the cursor fix-up).
No-region M-; is Emacs `comment-line`, not `comment-dwim`: toggle,
then move to the next line so repeated M-; walks a block (named
deviation; DWIM's append-at-EOL can come later under its own name).
Region toggles clear the selection and land at the span start. The
command boundary substrate provides chain-break and after-edit for
free (Q#CT6) — asserted anyway.
Tests (comment_toggle_acceptance, 14): rust/lua/python prefixes and
exact round-trips; cursor-next-line incl. the no-trailing-newline
clamp; region min-indent alignment + blank-line skip + selection
clear; mixed-span round-trip; region ending at column 0 excludes that
line; unknown-language and pathless-scratch no-ops; ONE undo restores
a multi-line toggle; rejecting/transforming intercepts (cursor fix-up
skipped); after-edit exactly once on both keybound and M-x paths;
C-k, M-;, C-k breaks the kill chain. Fixture editors empty
pmacs.lsp.config so .rs/.py files never spawn real servers.
Gates: fmt; workspace clippy -D warnings; lib 1500; crdt 1672;
comment 14; killring 30; cua 5; completion 9; autosave 29; m4 100
(--skip basedpyright); GPU 58 (PMACS_REQUIRE_GPU=1); full workspace
sweep clean; git diff --check clean.
Framing: docs/comment-toggle-framing.md.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MtRqijWecEzTjPt1B4Nrt5
Addresses the round-4 findings against the stack (PR #104 portion).
- HIGH range-only semantic-token servers: LSP defines
semanticTokensProvider.full and .range as optional, INDEPENDENT
capabilities, but the old any-provider gate sent /full regardless — a
range-only server rejects it and the swallowed error means no styling,
ever. Both the auto-pull and the manual command now gate each request
kind on its own capability: /full (delta under full.delta) when
negotiated; a range-only provider gets a WHOLE-DOCUMENT /range request.
New `rangeonly` fake mode (advertises range without full, rejects
/full) + test proving tokens arrive via the range path.
- MEDIUM completion acceptance left this_command stale: the popup accept
applies its edit and fires after-edit outside command dispatch, so
this_command could still read "buffer.self-insert" from the typing that
raised the popup — a candidate ending in "(" would spuriously
auto-trigger signature help. Accept now stamps its own boundary
("completion.accept"); asserted in the popup acceptance suite.
- MEDIUM GPU shape inference tightened: the 1-4-byte predicate accepted
a 2-byte "a(" insert (two ASCII codepoints). The classifier now decodes
the inserted bytes from the post-edit rope and requires the leading
byte's UTF-8 sequence length to equal inserted_len — exactly one
codepoint. The daemon unit test now drives an "a(" op and asserts it
breaks the chain instead of classifying as typing. Exact wire
provenance on the CRDT op remains the named deferred general fix.
Gates: fmt + workspace clippy clean; lib 1500; crdt 1672; m4 98;
completion 9; killring 30; GPU 58; git diff --check clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Addresses the four post-merge findings against PR #102 (merged as
2d157d8). Stacked on the kill-ring branch (PR #103): the trigger
redesign rides its command-boundary substrate.
- BLOCKING delta without the capability: pull_semantic_tokens_quiet (and
the pre-existing manual pmacs.lsp.semantic_tokens(), same bug) used
any stored resultId to request /full/delta while only checking that a
provider exists. A resultId does not imply delta support --- servers
may return one from /full regardless --- and a conforming full-only
server rejects the delta request; the pull path swallows the error, so
styling stayed silently stale after the first edit. Both sites now
require semanticTokensProvider.full.delta == true. The fake's default
mode truthfully advertises { "full": { "delta": true } } (it
implements delta); a new `fullonly` mode advertises "full": true,
REJECTS /full/delta, and bumps its resultId per /full response so the
test can observe WHICH pull refreshed the store. Verified the test
bites: with the capability check reverted, the post-edit rid stays
rid-1 (stale) and the test fails.
- HIGH false-positive trigger + cross-frontend misclassification: the
cursor-delta heuristic ("same buffer, cursor +1") fired on any
one-byte edit --- including a one-byte paste of "(" once PR #103 made
paste fire buffer.after-edit --- and its singleton last_typed was
shared across frontends. Replaced with the input-origin signal from
the #103 substrate: inside after-edit,
pmacs.editor.this_command() == "buffer.self-insert" names an edit
produced by typing, per frontend, with nothing inferred from cursor
deltas. New ed.this_command() binding; handle_remote_crdt_op now
classifies a single-codepoint optimistic insert as buffer.self-insert
(rotation, not just break --- kill-chain semantics identical since
self-insert is not a kill, and GPU typing now carries the same origin
signal as TUI typing). Paste/pointer/undo/unbound leave this_command
as something else and can never trigger.
- MEDIUM first-trigger-ignored: the origin signal needs no prior-edit
snapshot, so the very first "(" typed in a buffer triggers. The test
that had encoded the warm-up keystroke as "correct" now types a single
"(" as the first character.
- MEDIUM non-ASCII trigger characters: char_before read one byte and
rejected multi-byte strings; LSP trigger characters are strings. Now
codepoint-aware (read up to 4 bytes back, take the suffix from the
last non-continuation byte). The sighelp fake declares a two-byte
trigger ("«") and a test types it.
Tests (m4_acceptance 94 -> 97 after +4/-1 rework):
arc1c_full_only_server_repulls_via_full_not_delta (bites --- verified),
arc1d_signature_help_auto_triggers_on_trigger_char (now first-char),
arc1d_signature_help_triggers_on_non_ascii_trigger_char,
arc1d_signature_help_ignores_non_typed_edits (movement-stamped
programmatic "(" insert + manual after-edit must not trigger --- the
case cursor-delta inference cannot distinguish). Daemon unit test
updated for the insert classification (break-then-classify: `this` =
buffer.self-insert, `last` = None, chain still dead).
Note: completion.lua still uses the Q#C9 cursor-delta heuristic and
inherits its weaknesses; migrating it to this_command is a named
follow-up, out of scope here.
Gates: fmt + workspace clippy clean; lib 1500; crdt 1672; m4 97;
killring 28; completion 9; GPU 58; git diff --check clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Closes Arc 1 of docs/roadmap-2026-07.md.
1c --- semantic tokens never appeared (a shipped bug).
Semantic tokens are pull-model: the store only fills from a
`textDocument/semanticTokens/*` response. The ONLY automatic pull was in
reply to a server-initiated `workspace/semanticTokens/refresh`, which
most servers never send. So `LspStyleView` attached to a store nothing
ever filled, and semantic styling silently never appeared unless the user
ran `M-x lsp.semantic-tokens` by hand --- while inlay hints, on the exact
same pull model, were pulled at three points.
`pull_semantic_tokens_quiet` now mirrors `pull_inlay_hints_quiet` at all
three: on `initialized`, on attach, and on edit-flush. The `initialized`
handler is the one that matters --- buffers attach before the server
finishes initializing, so the attach-time pull is a no-op for the first
file (its `server_is_initialized` guard is false). That is precisely why
the file that starts the server never got semantic color. Delta when a
resultId is held, full otherwise, matching the manual command.
1d --- signature help auto-triggers on a trigger character.
A typed character is reconstructed the way `completion.lua` already does
(Q#C9): same buffer, cursor advanced by exactly one byte. Paste, undo,
kill, and remote CRDT edits produce any other delta and never trigger.
The trigger set comes from the server's declared `triggerCharacters` +
`retriggerCharacters`; a provider declaring neither gets `(` and `,`; no
provider means no auto-trigger at all. The request is silent --- an
auto-trigger that announced "no signature help" on every `(` in a comment
would be unusable --- so only a real signature reaches the status line.
It fires after the pending didChange is queued and flushes it first, so
the server sees the character being asked about.
Test helper: `pmacs_fake_lsp` gains a `sighelp` mode that advertises
`signatureHelpProvider`; every other mode omits it, so no existing test
changes behavior.
Tests (m4_acceptance 90 -> 94):
arc1c_semantic_tokens_auto_pull_on_attach (default fake: advertises
the provider, never sends refresh --- exactly the broken case)
arc1c_semantic_tokens_repull_after_edit_flush (clear store, type, flush)
arc1d_signature_help_auto_triggers_on_trigger_char
arc1d_signature_help_does_not_trigger_on_ordinary_typing
Verified the 1c tests bite: both fail with the `initialized`-handler pull
reverted. Named `arc1c_`/`arc1d_` rather than `m4_NN_`, since the m4
numbering maps to spec acceptance bullets and these are not those.
Gates: fmt + workspace clippy clean; lib 1499; m4 94; m9_1 18;
completion 9; listview 6; overlay 2; GPU 58; git diff --check clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Three PR #96 review findings (documentation accuracy):
- Q#P7 coordinates section claimed panels inherit a byte==UTF-16 wire
assumption with position-encoding hardening deferred. False as-built:
the transport layer negotiates general.positionEncoding and converts
every Position at the request/response boundary (PositionEncoding +
char_to_byte/byte_to_char, src/lsp.rs), so location rows reach Lua as
byte offsets. Reworded to record what landed; the true residual is
the codepoint-vs-byte cursor walk in move_active_cursor_to (shared
with go_to_definition, not introduced by panels).
- Intro described pre-arc behavior in present tense (references throw
rows away, code actions apply acts[1] blind, ...). Marked as the
pre-arc baseline with a status banner + inline as-built pointers.
- Drifted hard-coded line refs (editor_core.rs:2052-2071,
lsp.lua:658-662, lsp.lua:1187-1213) replaced with symbol names.
Also fixed the move_active_cursor_to comment in lsp.lua itself — it
was the same 'v0.2 hardening' false trail the doc's stale ref pointed
at, now naming the real residual (codepoint-walk, not wire encoding).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Pure Lua on the phase-1 substrate (framing Q#P5).
Outline: lsp.document-symbols (C-c o) opens *outline* -- the store's
FLAT symbol rows indent by their depth field with an LSP SymbolKind
tag; RET pushes the jump ring, restores the source buffer, and moves
to the symbol (M-, returns to the outline row, the references-panel
semantics).
Code actions: lsp.code-actions (C-c a) applies a single action
directly (previous behavior, now correct instead of lucky) and opens
the minibuffer dropdown when several are available -- 'N: title'
candidates; a bare typed index also accepts. The apply branch is
extracted as apply_code_action, shared by both paths. The m4_14/m4_15
acceptance tests (written against blind-first-apply; the fake LSP
returns two actions) now drive the picker: pump until the prompt is
live, type '1', RET -- same command-only action as before.
Hover doc: new lsp.hover-doc (C-c H) renders the full multi-line
hover contents into a non-visitable *lsp-help* panel; lsp.hover
(C-c h) keeps its one-line echo-area summary.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two PR #94 validation findings.
1. (High, blocking) GPU stuck after leaving a panel: the GPU only
swaps its displayed buffer on BufferSnapshot, and the daemon only
sent one on the first CRDT upgrade (F29's ensure returns None for
an already-backed buffer). A panel's q / RET switched the daemon's
active buffer back to the already-known source and sent nothing --
the GPU kept rendering the panel while input targeted the source: a
typing-into-a-buffer-you-can't-see hazard. Fix: the per-tick loop
now FOLLOWS each replica frontend's own active buffer -- when it
differs from the last snapshot sent to that frontend, ship that
buffer's snapshot to that frontend only (the F29 broadcast records
itself so the upgrade tick doesn't double-send). First-tick send
also repairs the attach-time last-snapshot-wins ambiguity. Snapshot
export extracted and shared with the F29 broadcast; per-fid state
cleaned on both detach paths.
2. (High, wider than reported) 'LSP doesn't activate on navigate':
switch_active_buffer clears the window's overlays, and the runtime
dedup tables (highlighted_buffers, styled_buffers,
diag_viewed_buffers) blocked re-attachment -- so EVERY buffer
switch (plain C-x b included, long-latent) permanently stripped
syntax color, LSP semantic style, and diagnostic underlines;
verified: overlay kinds [syntax-highlight, lsp-style, diagnostic]
-> [] after one away-and-back. Fix: a new additive
buffer.after-switch hook, fired by the window.switch_buffer binding
and find_or_open's existing-buffer branch; syntax.lua and lsp.lua
subscribe and re-push their views (the just-cleared window makes
that exactly-once per switch; fresh loads keep firing after-load).
Regression: tests/overlay_reattach_acceptance.rs (double round-trip
counts exactly one highlight overlay; panel q restores styling).
The daemon follow path is validated live (daemon + GPU) -- its unit
seam is the shared export helper.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Arc 1b phase 1 (framing: docs/lsp-panels-framing.md).
Q#P6 (the one Rust change): EditorCore.round_trip_buffers +
pmacs.buffer.set_round_trip_input(buf, on); dispatch_idle() reports
false while a marked buffer is active, so semantic frontends'
optimistic-apply stays off -- RET reaches a panel's buffer-local visit
binding instead of locally inserting a newline, and typing dispatches
into the edit path where the read-only intercept rejects it (a CRDT
import would bypass the intercept chain entirely). Pruned on kill.
Q#P1/P2/P3: builtin/runtime/listview.lua generalizes the *buffer-list*
idiom -- pmacs.listview.open{name, header, rows, on_visit, on_refresh}
owns ensure-buffer (recreates if user-killed), wholesale render with
bypass_intercept, line->item map, buffer-local RET/SPC/n/p/g/q keymap,
previous-buffer capture + q restore (never another panel; scratch
fallback), cursor re-seat after render, the read-only intercept, and
the Q#P6 mark. Panels are buffers: both frontends render them with
zero protocol change.
Q#P4: lsp.find-references (M-?) opens *references* -- one row per
location, paths shortened against the project root, RET visits via the
shared SP-4 template (jump ring, find_or_open, cursor walk; extracted
as visit_location for the phase-2 outline to reuse).
Acceptance: tests/listview_acceptance.rs -- open/seat/visit, header
non-visitable, q restore, read-only rejection, dispatch_idle gate,
refresh re-render + re-seat.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Five findings from the manual validation pass, all in-branch:
1. LSP-only words never queried the server: the auto-open path fired
request_completion only when the sync providers already produced
rows. An empty sweep now leaves a pending session and the request
always fires; isIncomplete responses re-request on further typing.
Corollary: attachment_for_request now flushes-if-attached but NEVER
attaches -- the first cut wrapped attached_for_active, which spawns
servers on demand, i.e. per-keystroke spawn attempts in unattached
buffers (wedged the parallel m4 suite; serial ran 3x slower).
Attachment stays buffer-open policy.
2. Cross-buffer LSP leak: the built-in provider's no-uri fallback was
the legacy global store drain, so scratch/unattached buffers could
show another file's cached completions. Strict now: no uri, no rows.
3. Pending prefixes own the keyboard: Action::Pending (C-x ...)
dismisses the popup and the popup shadow is guarded on an empty
dispatcher prefix, so the sequence's continuation and its C-g abort
reach the dispatcher instead of the popup.
4. Window-scoped sessions: CompletionPopupState.window_id (stamped by
completion_popup_open; Lua never sees it). Only the owning window's
overlay paints -- same-buffer splits each carry a persistent
overlay -- and a focus change invalidates the session.
5. Flaky worker test: the /proc thread-count probe and the idempotence
check both build EditorStates and could run concurrently, polluting
the baseline; merged into one test (non-Linux keeps a portable
idempotence variant).
Regression tests for 1-4; framing doc gains the as-built notes.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Q#C1/C9: builtin/runtime/completion.lua reconstructs typing intent
from state (buffer.after-edit has no payload): a {buffer, cursor}
snapshot recognizes the single-byte-advance typing signature, so
paste/undo/kill/remote edits never auto-open; prefix >= 2 opens off
the synchronous providers, server trigger chars open a pending session
that materializes when the LSP answer lands; refresh-on-typing
re-derives the prefix from the text; a core-closed popup suppresses
reopen off the same edit (the accept case). completion.at-point on
C-M-i covers deliberate invocation; the driver filters collect() to
score >= 0 (collect keeps non-matches, merely sorted last).
Q#C8: CompletionContext gains uri; the built-in LSP provider scopes to
it (legacy global drain only when absent); Lua providers get uri as a
trailing ninth positional arg; context_for can now express char
triggers + uri. pmacs.lsp.attachment_for_request() exposes the
flushing accessor (attached_for_active) so completion requests answer
against current text, not the debounced didChange backlog.
Q#C2 write path: pmacs.completion.popup_show/popup_hide/popup_visible
publish into the core session (kind tags shared with collect() rows,
so driver code passes rows straight through).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The Lua glue that gives the menu real items to surface.
- `edit.copy/cut/paste/select-all` commands (Q#CM6) over the core
clipboard, with the Emacs kill/yank bindings `M-w`/`C-w`/`C-y` and
`C-x h` (the CUA trio's keys are already bound: `C-a` line-start,
`C-v` page-down).
- `pmacs.lsp.active_attachment()` (Q#CM5): a pure, side-effect-free
attachment lookup for the menu's `symbol`/`diagnostic` visibility
checks. Unlike `attached_for_active`, it never triggers an attach just
because the menu opened.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014TXbAwk27agwhrNNrhLi2U
Full-document didChange went out per keystroke: three O(file) copies,
O(file) JSON, and a BLOCKING pipe write on the daemon main thread
(Linux pipe buffers are 64KiB; a 240KB notification stalls the frame
loop until the langserver drains). The dominant daemon-side typing
cost on large files, and freeze-class when a server stops reading.
- lsp.lua: the after-edit hook now bumps the version, marks the
cached render families stale (new _mark_document_stale binding, so
stale suppression stays keystroke-accurate), and records the buffer
dirty. The coalesced send fires on the async tick after 75ms of
quiet, or at most 400ms behind during continuous typing. Anything
that consults the server flushes first (attached_for_active,
repull_for_attachments, pull_inlay_hints_quiet) so requests and
position-encoding conversion never see stale text. Versions may
skip values; LSP only requires they increase.
- Inlay hints re-pull at flush cadence: they're pull-model, nothing
re-requested them after edits, so hints died on the first
keystroke and never returned.
- process.rs StdinWriter: a per-generation writer thread owns the
child's stdin; write_stdin queues and never blocks (64MiB budget
converts a wedged child into an error); close_stdin drains then
EOFs, preserving the MCP flush-then-EOF contract.
- pmacs.editor.monotonic_ms + pmacs.lsp._flush_did_changes bindings;
acceptance test pins burst-coalescing, flush-on-demand, and the
quiet-window tick flush.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Adds diagnostic navigation to the TUI/editor surface. Reuses the
existing `pmacs.diag.next` / `previous` walkers (which already wrap
around) and the cross-file jump ring so `M-,` returns from a
diagnostic jump just like an LSP definition jump.
Surface:
* `pmacs.command.define { name = "diag.next" / "diag.previous" }`
* `pmacs.keymap.bind { sequence = "M-g n" / "M-g p" }` — Emacs's
`next-error` / `previous-error` chord.
The command walks the diag store for the active buffer's attached URI,
falls back to a status-line message ("no LSP server" / "no diagnostics
in buffer") rather than faulting when there's nothing to jump to. On a
hit it pushes the jump ring, moves the cursor via `pmacs.editor` motion
primitives (so every overlay observer sees the navigation), and sets a
status line of the form `diag (warning): ...`.
Test verifies the commands are registered, bindings exist, and the
no-server status path lands.
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
The TUI's `DiagnosticView` has existed in `src/diag.rs` since v0.1 but
was never instantiated, so the local-grid renderer never painted
diagnostic underlines. This wires the view in the same way
`LspStyleView` and `SyntaxHighlightView` are wired — a Lua binding
that pushes the overlay onto the active window, driven from
`lsp.lua`'s `attach_buffer` flow with the standard per-buffer dedup
table.
* `DiagnosticView::kind()` returns `"diagnostic"` so
`pmacs.window._overlay_kinds()` can verify attachment.
* `pmacs.diag._attach_view(buf, uri)` mirrors `pmacs.lsp._attach_style`
exactly: requires active window's buffer matches `buf`, constructs
`DiagnosticView::new(uri, store)`, pushes as overlay.
* `lsp.lua` calls `pmacs.diag._attach_view` from `attach_buffer` and
tracks pushed buffers in `diag_viewed_buffers` to prevent
double-attach on repeated `attach_buffer` calls.
Scope is intentionally narrow: view attachment only. Navigation
bindings, statusline summary, and gutter signs remain follow-ups
under task #23.
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Drops the policy-A exclusivity that left grammar-backed languages
without LSP semantic refinement. Adds tree-sitter-c (.c/.h) and
tree-sitter-cpp (.cpp/.cc/.cxx/.hpp/...) to the bundle so the grid
TUI gets lexical highlighting (keywords / strings / operators) on
first open. The Lua attach in builtin/runtime/lsp.lua now pushes
LspStyleView whenever an LSP server is up, regardless of grammar
presence; with both views attached the cell-painter pipeline runs
SyntaxHighlightView first (lexical) then LspStyleView (semantic)
and their styles compose through crate::overlay::merge_styles. The
result is the VSCode / Zed "TextMate + LSP semantic tokens" model
on a terminal grid: keywords colored by tree-sitter, identifiers
refined by clangd's semantic tokens.
`.h` is ambiguous C / C++; the `c` BUILTIN_LANGUAGES entry claims it
to match the LSP filetype map's default. Users who want `.h` parsed
as C++ can override via Lua (extension → language map).
Note the tree-sitter-c / -cpp crates expose `HIGHLIGHT_QUERY`
(singular), matching tree-sitter-md's `HIGHLIGHT_QUERY_BLOCK`
convention; tree-sitter-rust / -lua use `HIGHLIGHTS_QUERY` (plural).
Same bundled highlights.scm either way.
Regression guard: builtin_languages_include_c_and_cpp asserts the
language entries exist and claim their canonical extensions. The
LspStyleView module doc rewritten to reflect dual-authority
composition; the existing headline test's comment updated (the
test fixture still attaches only LspStyleView directly, so its
asserted cells reflect the LSP authority alone — Lua-level
attach_buffer is what exercises composition end-to-end).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closes the visible "C++ has no syntax coloring in the grid TUI" gap.
Sibling of SyntaxHighlightView: a View impl that paints LSP semantic
tokens as cell styles, attached for buffers with no bundled
tree-sitter grammar. Same policy A (one styling authority per buffer)
the semantic-frontend producer arc enforces, applied to the grid
renderer the user actually uses today.
Mechanics: every render re-derives the buffer's URI from
buf.file_path() and pulls (encoding, legend) via the existing
LspManager::semantic_style_context plus tokens via for_uri. Per
visible line, tokens are converted from LSP encoding units to byte
ranges via char_to_byte, then to display columns via the existing
byte_range_to_display_cols (UTF-8 + tab aware). Theme::lookup
resolves token type names through the same dotted-prefix mechanism
the tree-sitter capture names use, so "function", "variable",
"type", "keyword" land on the existing theme vocabulary with no new
style names. Default-styled spans skip the per-cell loop, matching
SyntaxHighlightView's short-circuit.
Wiring: pmacs.lsp._attach_style binding pushes the overlay on the
active window (mirrors pmacs.parse._attach_highlight). install_lsp
and make_lsp_manager take SharedSyntaxRegistry so the binding can
hand the LspStyleView the shared ThemeHandle; editor.rs caller
updated. builtin/runtime/lsp.lua's attach_buffer attaches the view
when pmacs.parse.language_for_path returns nil (grammar-less
signal), dedup'd via a styled_buffers set that mirrors syntax.lua's
highlighted_buffers.
Test: lsp_style_view_paints_cells_from_semantic_tokens — seeds an
Initialized fake LSP client (using the cfg(test) helper from the
producer arc) on a /tmp/x.cpp buffer with one token, asserts the
expected cells are styled per the theme face and the cell just past
the token range is unchanged.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The default-bundle auto-attach path (lsp.lua ensure_server) never
forwarded cwd/root_uri to pmacs.lsp.spawn, so build_initialize fell
back to std::env::current_dir() — every auto-attached server received
the *editor's* cwd as rootUri regardless of which project the opened
file belonged to. Module-strict servers (gopls, rust-analyzer) return
nothing unless launched from the project dir; the fake-LSP and clangd
(which finds compile_flags near the file) masked this, gopls exposes
it. Same shape as the #26 transport bugs: lenient fakes hid a gap
strict real servers fall straight into.
Fix: project_root_for(language, path) in lsp.lua —
config[lang].root override -> pmacs.project.detect marker walk (the
canonical detector, honors set_search_boundary) -> the file's own
directory. attach_buffer resolves the path before ensure_server;
spawn now carries cwd/root_uri. Single-root only (fixes which root
the one per-language server uses); one-server-per-root multi-root
scoping stays deferred post-v0.1 (documented: first file of a
language fixes that server's root). New documented
pmacs.lsp.config[lang].root key.
Tests:
- m4_26: deterministic — new fake "rooturi" mode +
PMACS_FAKE_LSP_ROOT_SINK side-channel; asserts the rootUri sent
through a real find_or_open auto-attach is the go.mod dir, not the
cwd, not the file's own dir.
- m4_27: PATH-gated real gopls — documentSymbol + hover round-trip is
end-to-end proof of the fix against a real strict server.
- m4_28: PATH-gated real clangd — diagnostics arriving is the #26
deferred-notification-flush + URI-absolutization regression guard;
also exercises semantic tokens + documentSymbol.
No other latent bugs surfaced; gopls & clangd both clean through the
fixed path. rust-analyzer / basedpyright not installed here, so their
real end-to-end validation is still pending (the fix benefits them
identically — Cargo.toml / pyproject.toml are detect markers).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Ship single-binary LSP servers pre-wired in the default bundle so a
user who installs the server gets attachment with no init.lua:
- typescript-language-server (--stdio) for the typescript /
typescriptreact / javascript / javascriptreact language ids
- lua-language-server (settings.Lua present-not-null for the
workspace/configuration pull)
- bash-language-server (start subcommand)
- taplo (lsp stdio; settings.taplo present-not-null)
- zls (no args)
Plus the pmacs.lsp.filetypes extension->language map entries
(ts/mts/cts, tsx, js/mjs/cjs, jsx, sh, bash, toml, zig, zon, lua),
keeping the same idempotent `or` guard so init.lua overrides win.
m4_25 asserts every config table and the filetype map resolve to
the documented values (binary-independent, spawns nothing).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Dynamic file-watch registration with a full snapshot-diff watcher.
- src/lsp.rs: did_change_watched_files(sid, &changes) notification;
capability workspace.didChangeWatchedFiles.dynamicRegistration=true
(mandatory — clangd/rust-analyzer/gopls only register dynamically).
- src/lua_bindings.rs: pmacs.lsp.did_change_watched_files binding.
- builtin/runtime/lsp.lua: client/(un)registerCapability handled in
the server-request pump (reply null; start/stop watchers). Brace-
expanding glob → anchored Lua pattern; recursive read_dir/stat
snapshot-diff poller emitting per-file created/changed/deleted
filtered by glob + WatchKind, batched into one notification;
self-cancels when the server dies or unregisters. luajit-safe
(kind_has() arithmetic, no 5.3 bitwise).
- pmacs_fake_lsp.rs: `filewatch` mode registers a **/*.txt watcher
and logs received changes to <base>/.received (disk side-channel —
the protocol stream is drained by the pump).
- tests/m4_acceptance.rs: m4_24 asserts create(1)/change(2)/
delete(3) for matching .txt only; non-matching .md filtered.
Bug caught in validation: `**/` → `(.*/)?` is not a valid Lua
pattern (no group quantifier) — matched nothing, zero events. Fixed
to `**/`→`.-`, `**`→`.*`; m4_24 surfaced it.
client/unregisterCapability cancels watcher records (code-reviewed);
not asserted in m4_24 — a "no further notifications" negative-timing
check is flaky; the create/change/delete + filter path is the
deterministic proof.
Gates: lib 1301/0, m4 79/0, m8_1 10/0, m8_9 26/0, m8_10 19/0,
m9_1 18/0, m5_8 5/0, m11_5 (--features crdt) 2/0; fmt + clippy
clean.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Levi Neuwirth