Capture the managed daemon's lifecycle facts from the ready report before
waiting for the post-quiescence directory snapshot. If that regression
assertion times out, the acceptance harness can now terminate the daemon
instead of leaving it orphaned.
Drive pmacs --gpu . through the root broker and real managed GPU
connector, keep the session alive through the asynchronous dired
replacement, and assert its canonical listing before daemon reuse.
Expose snapshot count and materialized text through the private
display-less acceptance probe so the public path is observable.
Consume the directory session's later replacement snapshot and assert
the canonical dired header plus a known listing entry before checking
daemon reuse.
Correct the bottom-panel revision-5 recovery branch and advance the
durable handoff to the Journey Stage 1a main anchor.
Replace the stale directory-negative in the GPU initial-target suite
with an explicit snapshot-first readiness path, while retaining all
genuinely malformed and unloadable failure cases. Record the portable
side-quest and bottom-panel dependency state.
Review round 1 of PR #182. One implementation gap and two stale claims.
**The scope pins the frontend; it does not pin the window.** Framing
§4.4 specified `display{ window = dest:window() }`, but dired's commit
still ended in `pmacs.window.switch_buffer`, which targets whatever
window the scoped frontend has selected. A split or panel that took
focus while `read_dir` was pending therefore received the listing, and
`prev` was captured from it too — with every preflight check passing,
because the captured window was still live and still held its captured
buffer. Both sites now read the captured window: `display` routes to it
with `select = true` (the later `seat_cursor` acts on the active
window), and the `prev` read asks it directly.
N4c pins both halves. The suite's existing routing pins all varied
*frontend* identity; none varied the selected window within one
frontend, which is exactly why 23 green pins missed this. Bite: dired's
`display` back to `switch_buffer` fails N4c alone; `prev` read from the
ambient window fails N4c alone.
Two stale documentation claims, both of which this PR was supposed to
have already fixed:
* **The §0 scorecard still graded §2 "Broken at entry"** while §2's own
ground truth had been rewritten. The scorecard is a second copy of the
same claim and §25's protocol covers both. §19's row and ground truth
were stale the same way — this PR creates the first cross-subsystem
suite, which §19 says should exist and grades as missing — and are
corrected too.
* **P4 still read "leaves exactly one buffer"**, the exact claim rev 6
corrected as false everywhere else in the framing. Restated to what it
actually pins: the file is in the *active window*. The test was
already written correctly; only the framing lied.
Framing rev 8.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Completes Journey Stage 1a: the `commit_to` acceptance suite (framing
§6 N4, N6, N6b, N6c, P1, P2, P3) plus the documentation updates
COHERENCE §25 requires the PR to carry.
Bite-testing the new pins found a real gap. Deleting the
`ScopedFrontend` arm from `acting_frontend` left N4 green, because
`ScopedFrontend::enter` also swaps `core.active_frontend` and the
ambient fallback then answers correctly on its own. The arm is
load-bearing in exactly one case — a commit reached from inside an
interactive command, where the origin sits between the override and the
ambient value — and nothing pinned it. N4b is added, driven through
`dispatch_key` because that is the only thing that establishes an
interactive origin, and the mutation now bites it.
Two smaller corrections found the same way:
* `commit_to`'s forged-destination message was unreachable. Typed as
`AnyUserData`, mlua rejected a table during argument conversion, so a
caller got "error converting Lua table to userdata" — true, but naming
neither the rule nor the remedy. The parameter is now `mlua::Value`
and the pointed message fires.
* P1 and P2 also fail on full revert, since `commit_to` does not exist
on the pre-image, so §6.0's "legitimately green on the pre-image" does
not describe them. They stay in the P list because their
discriminating falsifier is the named mutation — a revert-only check
cannot distinguish "validates" from "validates in time" — and each pin
now says so at its own site rather than being silently mislabelled.
Bite results, each run against the whole suite:
scope stops swapping `core.active_frontend` -> N6a, P3 fail; nothing else
preflight moved after the callback -> P1, P2 fail; nothing else
drop the `ScopedFrontend` arm -> N4b fails; nothing else
Docs: COHERENCE §2 grade + step-3 verdict row, §20 Priority 1 and the
arc list; the GPU initial-target framing's Q#GT6 and acceptance 10,
whose directory case this stage deliberately supersedes; handoff §1;
the active-work ledger; framing rev 7.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Round 11 put the nesting count in the expander, which is optional. A
consumer at a lower priority can CLAIM and stop the chain before the
expander runs, while that fan-out's deferred-expansion subscriber still
runs — so the nested pass went uncounted, looked like the outermost
one, expanded early, and outer pairing resumed with a record the
replace had invalidated. `\alp(` gave `α(` again.
The count now comes from a no-op consumer registered at the minimum
priority, which runs first in every chain invocation that reaches any
consumer at all. Its guarantee is exactly the ordering contract the
chain already rests on, and it degrades safely: the only thing that can
skip it is a claim ahead of it, which skips the expander too, so
nothing is queued in that fan-out either.
The other plausible home does not work and the comment now says why: a
subscriber registered beside `run_deferred` is too late, because the
whole nested fan-out completes inside the OUTER chain's subscriber,
before either of them runs.
Acceptance 45o pins the short-circuit path — a consumer at 25 that
claims when the record is nil, so the nested pass never reaches the
expander. 45n passes against this bug, which is why both exist.
Counting in the expander fails 45o and nothing else.
Framing rev 12 also names the shape rounds 10–12 share: each fix was
correct about the failure it was shown and wrong about the boundary of
the mechanism it leaned on — the chain's copy semantics, then its
re-entrancy, then its short-circuit. A queue that outlives the thing
that filled it has to name that thing, not approximate it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011LFvC4FQtux4y32KuevZ7B
`buffer.after-edit` fan-outs NEST — the typed-edit contract supports a
consumer calling `pmacs.hook.run`, and typed_edit.lua's header says so
in its second paragraph. A nested run re-enters every subscriber,
including the deferred expansion's, while the OUTER chain is still
walking its consumer list and pairing has not yet seen the terminator.
So a consumer registered at priority 75 — between the expander at 50
and pairing at 100 — that runs one nested fan-out made `\alp(` yield
`α(` again: the nested pass consumed the queued expansion and edited,
and outer pairing then resumed holding a record the replace had
invalidated. That is round 10's failure reached through the chain's
documented re-entrancy seam rather than through claiming, which is why
deferring alone did not close it.
Deferring work past a fan-out means owning WHICH fan-out it belongs to.
The chain's subscriber and this module's each run exactly once per
fan-out, in that order, so counting invocations of the first and
matching them off in the second identifies the nesting level. Only the
outermost pass expands; a nested one leaves the expansion queued. No
new seam in typed_edit.lua, which is merged Stage 4a substrate.
Both halves bite: removing the level check and never counting
invocations each fail the new acceptance 45n.
Also fixes a test comment that still described the span design round 10
discarded — it claimed the expansion replaces the span "INCLUDING the
terminator". The behaviour asserted was right; the explanation was
stale. Framing rev 11.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011LFvC4FQtux4y32KuevZ7B
Three defects, all about what happens AROUND the expansion rather than
about resolving an abbreviation.
A pair character that TERMINATES an abbreviation never reached
auto-pairing: `\alp(` gave `α(`. Q#LN22 already said the terminator is
not claimed and the implementation claimed it whenever an expansion
succeeded. Merely declining is not enough either — the chain hands each
consumer a copy of the record made before any consumer ran, so
expanding inside the chain invalidates the copy pairing is holding and
the closer is silently lost. Verified by mutation rather than assumed:
expand-then-decline reproduces `α(` exactly.
The expansion therefore runs on its OWN `buffer.after-edit` subscriber,
registered after typed_edit.lua's and before lsp.lua's. A claim stops
the chain but not a separate subscriber, which is the point: pairing
claims the terminator it reacts to. The replaced span now covers only
the leader and the typed text, so pairing's closer lands outside it and
survives. One undo restores the same text either way, because the
terminator was always its own insert.
That second subscriber is a new instance of Q#AP7 — lsp.lua flushes
didChange synchronously on the signature-trigger path, and `(` is a
trigger — so acceptance 45m pins it with the sighelp fake server: no
didChange may ever carry the unexpanded text.
The relevance check is now three-part, as pairing's has been since
#110: buffer, window, AND `ed.cursor() == rec.post_cursor`. A redefined
self-insert can insert the completing character and then move the
point, and expanding over a span the user has left teleports them back
into it.
Cursor placement after the replace is context-guarded, as
`repair_cursor` is. A buffer intercept may switch buffers while
`buf:replace` runs; the unguarded `goto_byte` then translated the Lean
buffer's pre-edit point through the Lean buffer's edit and applied it
to whatever was ambient.
Q#LN22, criterion 38's span wording, and the ledger are corrected to
describe the deferred design rather than the one that shipped — the
rationale's source, not only the sites quoting it. Acceptance 45j/45k/
45l/45m added; framing rev 10.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011LFvC4FQtux4y32KuevZ7B
Four review findings, all confirmed against the tree.
Q#JR3 was false. `replace_active_buffer` does not drop the startup
scratch buffer -- its body is one `switch_active_buffer` call, which
reassigns the window's buffer_id and removes nothing. The claim came
from that function's own doc comment, wrong for as long as it has
existed, and rev 5 propagated it into the framing and into new
documentation this branch added. Both comments are corrected here,
because this PR was adding further false references to a claim P4
depends on. Actually removing the stale scratch is buffer-lifetime work
and stays out.
The daemon bootstrap could report the wrong buffer. The directory arm
captured the destination id, ran the resolver chain synchronously, then
returned the captured id -- so a handler that opened something
synchronously through commit_to had already replaced the window's
buffer, and the reply paired one buffer's snapshot with another's
identity. It also returned early, skipping the post-hook revalidation
the framing said stayed active. The arm now re-reads the destination
after dispatch and rehomes through `non_side_target` as the file arm
does. Pinned by a test whose handler claims synchronously.
N11 tested neither RET nor self-insert: it called display_file and
buf:insert directly, so it stayed green with dired's RET binding, its
entry dispatch, and the editor's self-insert path all broken. Both
gestures now go through dispatch_key.
P7 is removed rather than weakened. Q#JR12 has nothing to pin --
`had_file = file.is_some()` and a directory is Some like any other, so
no directory-specific branch exists to break. The old test never armed
restore and hard-coded had_file, so it could not fail against any
implementation.
Also adds the daemon bootstrap pins (N2, N5) and fixes an insertion that
had orphaned a `#[cfg(feature = "crdt")]` from the test it guarded --
which would have made one new test dark and one existing test escape its
gate.
Framing: docs/journey-stage1a-framing.md rev 6.
Journey Stage 1a's core: `pmacs .` opens the directory instead of
exiting 1, and local startup stops being a second implementation of
path resolution.
`EditorState::open` now calls `EditorCore::resolve_target_buffer` --
the primitive whose own doc comment says it exists "so two
path-normalization, dedup, and hook transactions cannot drift apart",
and which local startup had never been a caller of.
`resolve_target_buffer` returns a typed `ResolvedTarget` rather than
`(BufferId, HookKind)`, with a `Directory` arm checked ahead of the
load. Without it the load runs and fails: `File::open` succeeds on a
directory and `read_to_end` returns EISDIR, which is not `NotFound`, so
the `[new file]` arm never fired.
A directory creates no buffer. It dispatches a resolver chain: the
short-circuit `path.open-directory` hook, which no builtin subscribes
to, and then `pmacs.path.directory_handler`, which dired defaults. The
split is forced rather than chosen -- hook callbacks only append and
builtins load before init.lua, so a subscribing builtin would always
claim before any user listener could run. A raising listener stops the
chain and suppresses the fallback.
The listing is async and the daemon bootstrap is not, so the whole
post-await commit runs inside a new `pmacs.window.commit_to`: it
validates the destination -- frontend live, window live, buffer
unchanged, window replaceable -- BEFORE invoking its callback, then
scopes the acting frontend for its extent. Validating at display time
would be four dired mutations too late.
That scope is deliberately not `InteractiveCommandOrigin`, which does
not reach the core-ambient APIs and is authenticated user-command
authority a background continuation must not acquire.
The dedication rule is extracted into one `window_accepts_buffer`
shared by exact display, the display probe, and the new preflight, with
`incoming: Option<BufferId>` -- `None` means "the replacement does not
exist yet" and refuses a dedicated window.
`display_file` keeps its directory-is-an-error contract and does not
enter the chain; find-file's accept arm depends on it.
Framing: docs/journey-stage1a-framing.md rev 5 (Q#JR1-JR15).
Reconciles the handoff and ledger against a main that advanced past
this branch's base: the header, `main` anchor, and canonical-base
description take main's richer versions restamped to 74301d1, main's
new PTY-terminate lane is kept alongside the Lean lane, and main's
Stage 4a/rev-8 lane history is dropped in favour of the Stage 4b lane
that supersedes it — per this ledger's own rule to remove entries when
their PR merges.
Also fixes the coherence census's second count, which still said eight
settings three paragraphs below the nine it now lists.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011LFvC4FQtux4y32KuevZ7B
Typing `\alpha` in a Lean 4 buffer gives `α`; `\<>` gives `⟨⟩` with the
point between them. The abbreviation table is vendored from
vscode-lean4 and the expander is a typed-edit consumer registered on
the Stage 4a chain at priority 50, ahead of auto-pairing.
The ordering is load-bearing. 64 abbreviation keys contain a character
in the `lean4` pair set, so with pairing first, typing `\[` would
insert `[]` and corrupt the pending key to `\[]` before the second `[`
arrives — `\[[]]` becomes unreachable. The consumer therefore claims
every keystroke that EXTENDS a pending abbreviation, not only one that
completes an expansion; claiming only completions would hand each
intermediate `[` to pairing by a different route.
The vendored table is an ORDERED SEQUENCE, not a map. Upstream breaks
equal-length ties by source declaration order — 101 prefixes depend on
it, and `\f` resolves through `f<` rather than `f>` — which a
`pairs`-iterated Lua table cannot express. `scripts/regen-lean-abbrev`
takes a vscode-lean4 commit, emits the file with its provenance header,
and aborts on a duplicate key, invalid UTF-8, or a round-trip mismatch.
Undo is cross-peer-degraded on CRDT frontends and that is accepted and
named, not papered over (Q#LN21): `\alpha` arrives as six source-peer
optimistic inserts while the expansion is one daemon-peer replace.
`set_round_trip_input` would fix it and also makes `dispatch_idle`
report false, so RET would stop inserting a newline.
Round 9 corrects three approved acceptance criteria that the real table
contradicts, found by simulating the state machine over all 1,855
entries and re-reading upstream at the pinned commit rather than
re-reading the prose. `\to` is not eager — `top`, `to0` and `toa`
extend it. `\zzzz` expands to `ζzzz ` because `ze`, `zeta` and
`zsqrtd` exist; only `$ % , ; @ W` open no key at all. And `\alpha`'s
undo does not restore `\alpha ` because `alpha` IS eager, so the
terminator is a separate edit. Criteria 38, 41 and 42 now state both
paths, and the false halves are asserted too: they read as correct
until the table is consulted.
Three implementation traps worth the record. The generator's own
round-trip check was broken twice and failed closed both times:
`str.splitlines()` splits on U+2028, which 53 symbols contain, and
escaping through `chr(byte)` produced a latin-1-shaped string that the
UTF-8 write re-encoded. The first check compared in-memory strings and
agreed with itself; it now stages the file, re-reads the bytes from
disk, and renames into place only on a match. And the expansion SHRINKS
the buffer, so the point must be placed explicitly — pairing's
no-cursor-motion rule holds only for an insert AT the cursor, and
without this every self-insert after the first expansion is silently
rejected and the editor looks dead.
25 acceptance tests plus one `--lib` test for the optimistic CRDT
producer (45f), which is where the gate list's `--features crdt` run
reaches it; a crdt-gated integration test would be dark in CI and in
the gates both. Fifteen mutations bite, each failing its target. Three
of these tests were vacuous when first written and biting is what
found them: the abandonment test asserted text a surviving record
would also produce, the re-arm test used an example that never reaches
the re-arm branch, and both switch tests ran through
`find_or_open`'s fresh-load path rather than `buffer.after-switch`.
No protocol change (Q#LN14). Also reconciles the handoff and ledger
for Stage 4a (#179) and adds `lean.abbrev` to COHERENCE.md's
config-registry adoption census, now nine settings.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011LFvC4FQtux4y32KuevZ7B
Review round 4, P2. A fix can invalidate a test that was never written.
Criterion 17 still specified the pre-round-2 world: remove
`set_round_trip_input` and the optimistic op "passes `ensure_writable()`
and mutates BOTH sides, silently, with no divergence to notice". That
was true while no Lua binding set `read_only`. Since
`set_generated_contents` does, the daemon refuses the op — so only the
frontend's own mirror mutates, and the copies diverge.
The gap matters precisely because 17 is unpinned. A real-GPU test
written to the old spec would hunt for a daemon-side edit that can no
longer occur and pass for the wrong reason, quietly readmitting the
round-2 regression through a test not yet built. The specification is
the artifact under review here, not the code.
Restated around unauthorized MIRROR mutation plus daemon refusal —
divergence — in all four places carrying the obsolete claim: the
criterion itself, the Q#TC6a heading, the acceptance-16 doc comment, and
the bite roster. The heading's "ONLY thing" now says what it is the only
thing FOR: the replica's own mirror. `docs/active-work.md` also still
described acceptance 16b as asserting `is_read_only()` is false, which
round 2 flipped.
Why round-trip input stays load-bearing rather than redundant, now
stated wherever the daemon guard is mentioned: a refusal arrives after
the frontend has already applied optimistically and painted. It buys
divergence instead of silent agreement; it does not prevent the mutation
the user is looking at.
Also recorded, after capturing it properly this time: the gate-run flake
in `cargo test --lib --features crdt` is
`process::tests::setsid_escapee_is_not_reaped_and_teardown_reclaims_readers`
(`active_reader_probe` -> None, "live runtime probe"), ~1 run in 5.
Pre-existing and unrelated — this branch does not touch
`src/process.rs`, the test passes 10/10 standalone and 2017/2017 at
`--test-threads=1`, and it is another instance of the known `drain_until`
trap: draining for `Started` also ticks, and a tick reaps the leader.
That also explains the unattributed "2 failed" run noted in round 2.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016gGQC6eqHJVbZJ5Hg7aLer
Review round 3, on the round-2 primitive itself. One lesson covers all
three findings: a rope write is only half of an edit, and "discard
history" means whichever history the buffer actually has.
P1 — the binding swallowed the edit. `set_generated_contents` returned
`()`, so nothing reached `notify_buffer_edit_to_windows`. Two
consequences, both reproduced by the reviewer. In the default build a
window showing the buffer kept a `TextView` line index describing the
PREVIOUS contents, and the next paint indexed the new rope with stale
ranges — `assertion failed: end <= self.len()` in `src/rope.rs`. In the
CRDT build `pending_crdt_ops` stayed empty, so replica mirrors never
imported the owner's write and their optimistic edits were generated
against content already replaced. The `delete`+`insert` pair this
replaced had done that fan-out for free.
Now applies ONE whole-buffer `Replace`, returns its `Edit`, and notifies
from the binding. The doc comment states the obligation, because the
next owner to adopt the primitive inherits it.
P2 — "discard history" was false in CRDT mode. The v0.1 stacks are
bypassed entirely there; the history lives in loro's `UndoManager`.
`read_only` stops the replay but not the retention, which is the memory
cost the contract claims to eliminate. `UndoManager` exposes no clear,
but needs none: it records only what happens after it is constructed,
the same property `CrdtState::from_bytes` already uses to keep the seed
insert out of undo. `CrdtState::clear_undo_history` rebinds a fresh
manager to the same doc.
P2 — the docs described the pre-fix architecture. Q#TC6a said no Lua
binding sets `read_only` and round-trip input is the only guard; the
acceptance text still said `is_read_only() == false` while 16b had been
flipped to true; `terminal.lua`'s comment repeated the obsolete claim.
The architecture is layered and now says so: rope-level read-only
protects the daemon copy, round-trip input protects the replica's
optimistic mirror, and neither substitutes for the other. Q#TC6a keeps
its analysis under a superseded-in-part box rather than being silently
rewritten — its conclusion survives, two of its premises do not.
New pins. acc16d paints the window after a SHRINKING generated write:
stale offsets then point past the buffer end, so the failure is the
reported crash rather than merely stale pixels. acc16e asserts the
refresh is queued for mirrors, through the real copy-mode path;
`crdt`-gated and therefore dark in CI, which is why 16d drives the
binding rather than the terminal. Plus a CRDT unit test that ten renders
leave the `UndoManager` with nothing recorded.
Bites: dropping the notify panics acc16d at `rope.rs:145` and fails
acc16e with `queued: []`; dropping the `UndoManager` rebind fails the
new unit test on `can_undo`.
Still open, and recorded in COHERENCE.md §14: the fan-out obligation
makes `*compilation*`/listview adoption more than a one-line swap.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016gGQC6eqHJVbZJ5Hg7aLer
Five defects in the chain itself, plus the stale handoff state.
Each consumer now gets its own shallow copy of the typed-edit record.
Handing everyone the same table let a DECLINING consumer rewrite
provenance for the ones behind it, and pairing decides what to close
from `rec.char` — so a forged `char` turned a typed `x` into `x)`.
Every field is a scalar or an opaque id, so a shallow copy is complete.
The fan-out iterates a snapshot of the consumer list. It was iterating
the same array `add_consumer` mutates: a consumer that registered a
lower-priority one shifted itself forward under `ipairs` and ran twice,
and re-registering made that unbounded. Registrations and removals made
during a fan-out now take effect on the next one, stated as a contract
and pinned in both directions.
`tostring` on the caught error moved inside the containment. A Lua
error may be any value, including a table whose `__tostring` throws —
rendering it outside the `pcall` reintroduced exactly the escape the
containment exists to prevent.
Priorities are validated as finite integers in i32 range, matching
`pmacs.completion.register`. NaN is a number and every ordered
comparison with it is false, so a NaN consumer landed wherever the
insertion scan gave up and silently voided the lowest-first ordering
that Q#LN22 depends on.
`add_consumer` returns a handle and `remove_consumer` unregisters it,
reporting whether it was live. Without teardown the chain inherited the
`pmacs.hook.add` callback leak COHERENCE.md §13 already records, and
spread it to every consumer.
Also corrects the rationale the containment was documented with, in the
module, the test, and the framing: an uncontained throw does NOT take
the fan-out's other subscribers down. `run_all_must_succeed`
(src/hook.rs:332) collects errors and continues, so lsp.lua still
flushes didChange. The containment is still required — the throw skips
every later consumer in the chain — but the reason is narrower than
rev 7 claimed.
Criteria 46f (record isolation), 46g (snapshot iteration), and 46h
(lifecycle and priority validation) added; 46d's rationale corrected.
Four new tests, all bite-verified by mutation, each failing only its
target: shared record table (1), live-array iteration (1), unprotected
tostring (1), bare number check (1), no-op removal (2). The suite also
runs green under `--features lua54`.
docs/agent-handoff.md said Stage 4a was awaiting approval while this
branch had it implemented and in review.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011LFvC4FQtux4y32KuevZ7B
**P1-1 — layout invalidation could suppress the authoritative clear.**
Real bug. Both render paths resolved the document identity AFTER the
evaluator ran callbacks, but BOTH outcome arms carry PHASE-1 contexts.
A provider that closes the primary document split changes
`primary_document_window` mid-evaluation, so the filter compared
phase-1 contexts against a replacement identity, matched nothing, and
emitted no clear — leaving stale statusline text on the wire forever.
The identity is now captured BEFORE `evaluate_statusline` runs and
threaded through both paths (the terminal path via `terminal_chrome`).
Pinning it took three attempts, and the two failures are the useful
part:
- `pmacs.window.close()` takes no argument — it closes the ACTIVE
window. The first version passed a window id that was silently
ignored, so it closed the panel instead of the document.
- The Lua window API acts on the ACTIVE FRONTEND, so driving it against
a synthetic semantic view changed nothing at all.
- Closing the only document window is structurally REFUSED (Q#BP6
forbids a lone side window as a resting state), so the fixture needs
TWO document windows for the close to be legal.
The test now asserts its own precondition — that the callback really
changed the identity — before asserting the clear, and reproduces the
reported symptom (no `StatuslineSegments` at all) when the fix is
reverted.
**P1-2 — #21 was pinned at the helper, not the producer.** Confirmed:
reverting only the call site inside
`publish_buffer_snapshot_to_replicas` left both the helper test and the
existing socket-pair test green. The helper assertions are removed (with
a note saying why) and replaced by
`snapshot_publication_follows_the_document_under_a_focused_panel`, which
drives the real producer over socket pairs and asserts BOTH directions:
the document buffer's snapshot is delivered while a panel holds focus,
and a panel-only buffer's is not.
Biting that test exposed a defect in the test itself: the delivery read
had no timeout, so a regression made it HANG rather than fail. A hanging
test is strictly worse than a red one — every read now has a timeout.
Gates: fmt clean; workspace clippy clean; 1,832 default + 2,015 CRDT
library; Stage 2A 17; Stage 1 46; statusline 8; m11_5 2; GPU initial
target 14; terminal config 12; folding Stage 2 48; vterm 1/2 10 / 6;
M4 121; required GPU 202; `git diff --check` clean.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Review round 2, P1. Undo could empty the "read-only" snapshot.
`render_snapshot` wrote with bypass_intercept, which leaves ordinary
undo history behind, and `Buffer::undo` reaches the rope through
`ensure_writable` without ever consulting the intercept chain. So a
single `C-/` — or `M-x buffer.undo`, which needs no keymap at all —
replaced a freshly rendered snapshot with an empty buffer.
`set_round_trip_input` does not help: it routes the key into the daemon
command path, which is exactly where undo runs.
Rebinding the undo chords buffer-locally would not have closed this,
and `compile.lua` already says so in a comment: "command/menu undo
stays dispatchable". `*compilation*` and listview panels therefore
carry the same latent defect today.
Adds `Buffer::set_generated_contents` (Lua:
`pmacs.buffer.set_generated_contents`): lift `read_only`, replace the
contents skipping intercepts, discard the resulting history, re-assert
`read_only`. This ships the framing's deferred immutability lane as ONE
primitive rather than exposing the setter — a bare `set_read_only`
would let a caller lock a buffer it can no longer refresh, which is
precisely why that lane was deferred. Discarding history is
load-bearing twice: it removes what undo would replay, and it stops a
periodically refreshed buffer accumulating rope clones that `read_only`
guarantees nothing can ever pop.
New acceptance 16c drives the real M-x path
(`command.invoke_interactive`), the chord, and redo, and asserts the
owner's own refresh still works — the operation plain `read_only` would
have broken. Acceptance 16b flips from asserting `is_read_only()` is
false to true, because the property it documented is the one that was
wrong. Three `buffer.rs` unit tests cover the primitive directly,
including that ten refreshes leave an empty undo stack.
Bite: restoring the delete+insert render reproduces the report exactly
— `left: Some("")` against the full snapshot — failing 16c and 16b.
Still open, and now named in the framing, COHERENCE.md §14 and the
ledger: `*compilation*` and listview have not adopted the primitive and
remain emptiable by `M-x buffer.undo`; a streaming-friendly variant is
needed for the append case. In CRDT mode `read_only` is what refuses
undo, since loro's UndoManager exposes no clear through `CrdtState`.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016gGQC6eqHJVbZJ5Hg7aLer
`pmacs.editor.take_typed_edit()` is one-shot and per-frontend (Q#AP9):
the first `buffer.after-edit` callback to call it clears the slot, and
every later callback in the same fan-out sees nil. That was survivable
only because auto-pairing was the sole consumer — never a property
anyone chose. A second independent caller would get nil or steal the
record from pairing depending on hook registration order, and
registration order is not a contract.
This makes it one. `builtin/runtime/typed_edit.lua` owns the single
after-edit subscriber that reads the record, and offers that one read to
consumers registered through `pmacs.typed_edit.add_consumer{ name,
priority, fn }`: lowest priority first, ties by registration order, and
the first consumer to return truthy claims the edit and stops the chain.
`pair.lua` becomes that chain's only consumer, at priority 100.
No Lean content. Stage 4b's abbreviation expander is what needs the
ordering guarantee (64 of its 1,855 keys contain a `lean4` pair-set
character, so pairing running first corrupts them), but the chain is
substrate every language runs through, which is why it ships alone —
framing Q#LN10, and §4's rule that no PR in this arc mixes a
cross-cutting substrate change with Lean feature content.
Three design points worth review attention:
- Consumers are called even when the record is nil. "This fan-out
carried no typed edit" is information a consumer acts on: it is how
pairing's test seam observes a non-event, and how Stage 4b will
abandon a pending abbreviation an unrelated edit invalidated. Three
existing auto-pairing tests fail if the chain skips consumers on nil.
- The chain pcalls each consumer. `buffer.after-edit` is
all-must-succeed, so a throwing consumer would otherwise fail the
fan-out for every other subscriber, including lsp.lua's didChange
flush. Behavior-preserving for pairing, which already never throws.
- Ordered insertion, not `table.sort`, which is not stable in Lua —
"ties by registration order" is a stated contract, not a coincidence.
`tests/auto_pair_acceptance.rs` is UNCHANGED — zero lines — and its 45
tests pass. That is criterion 46 and the whole no-behavior-change claim;
a suite edited to accommodate the refactor would prove nothing.
`tests/typed_edit_chain_acceptance.rs` adds 9 tests for criteria
46a-46e. Every one is bite-verified by mutation: appending instead of
ordered insert (5 fail), `>=` for the tiebreak (1), re-taking per
consumer (4), ignoring the claim (1), dropping the pcall (1), skipping
nil fan-outs (1 here plus 3 in the untouched auto-pair suite), and
loading the chain after lsp.lua (the Q#AP7 flush test fails, alongside
the two existing pairing ones).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011LFvC4FQtux4y32KuevZ7B
Four findings, all real, and they rhyme in pairs. Two implementation
defects and two vacuous pins, all four tracing to one root: a name is
not an identity, and a context-free readout is not a state observation.
A foreign buffer carrying the snapshot's name was adopted and then
overwritten. `pmacs.buffer.create` accepts any caller-chosen name and
snapshot writes use bypass_intercept, so found-by-name adoption
clobbered user data — the reviewer reproduced "do not clobber" becoming
23 newlines. Now follows dired's F7 rule: ownership means "in copy
mode's own handle table", never "found by name", and a taken name yields
a `<2>` variant.
Snapshot identity was keyed by terminal NAME.
`TerminalManager::open` uniquifies only the derived name — an explicit
`name = ...` is inserted verbatim — so two valid terminals can share
one, and a name-keyed table handed them a single snapshot: the second
invocation retargeted it, `q` returned to the wrong terminal, and
killing either removed the shared buffer. Identity is now the terminal
buffer, compared in an array, because BufferIdLua implements `__eq` but
each wrapper is a distinct table key: comparison works, hashing does
not. The kill-with-terminal callback now closes over its own record
rather than looking the name up again.
The refresh pins were vacuous. Acceptance 19 compared a quiet
terminal's snapshot against itself and 18 counted buffers, so both
passed with render_snapshot replaced by a no-op. The child is
`exec cat`, so the tests now type a marker into the focused terminal,
require it ABSENT from the existing snapshot, and only then refresh —
via `g` and via re-invocation respectively.
The tail-follow pin could not observe view state.
`TerminalManager::snapshot(buffer_id)` is context-free and always
returns the live screen, so it reported "at the tail" even for a view
forced to the oldest retained row. Now read through
`snapshot_for_view`'s at_bottom and its projected cells.
Adds acceptance 18a (a foreign same-named buffer is never adopted or
clobbered) and 18b (two same-named terminals get two independent
snapshots, each `q` returns to its own source, and killing one leaves
the other's snapshot alive).
Four new bites, all discriminating: restoring adopt-by-name fails 18a
AND 18b; restoring name-keyed identity fails 18b; making
render_snapshot a no-op fails BOTH 18 and 19, which is the vacuity
demonstrated rather than argued; and forcing the view off the tail
fails 20.
Criterion 17 stays a named follow-up, per review agreement, until the
real GPU probe is non-skipping and CI-executed.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016gGQC6eqHJVbZJ5Hg7aLer
**P1-1 — the `Invalidated` arm published the panel context on the
document wire.** Real bug, and the live half of the routing defect: the
semantic peer has ONE statusline slot, so emitting an
authoritative-empty payload for every context replaced the document's
with the panel's. Now filtered by document-window identity exactly like
the `Ready` arm; a panel's own clear belongs to `PanelFrame` in 2B.
Pinned by `invalidated_statusline_clears_only_the_document_not_the_panel`,
which reproduces the reported shape — two targets instead of one — when
the filter is removed.
Honest note on the `Ready` arm: its identity selector is **defensive**,
not independently falsifiable today, because the document context is
captured first so "first context for my frontend" happens to pick it.
Rather than leave that as a silent dependency,
`the_semantic_fan_out_captures_the_document_first` pins the order and
says why it matters.
**P1-2 — round-1 finding 3 was not closed; four of my pins were
vacuous.** All four confirmed and fixed:
- The statusline consumer test discarded `render_frame`'s output. It now
observes the WIRE payload from a v18 peer with a registered provider,
and asserts non-emptiness so it cannot pass by emitting nothing.
- The terminal test compared two NON-terminal buffers, so both routings
answered `false`. The document window now holds a REAL terminal, so
the routes disagree; reverting `semantic_terminal_key` fails it.
- The decorations test used different buffers and an empty selection —
again the same answer either way. The panel now displays the declared
buffer with a non-empty selection while the document has none.
- #1/#3/#21 had no discriminating pin at all. Their only production
caller is `dispatcher_loop`, which no test can drive, so this extracts
three named seams the loop calls — `document_buffer_to_follow`,
`document_cursor_byte`, `peer_displays_buffer_as_document` — and pins
each.
Also newly pinned: #2 the lazy CRDT upgrade (the census's sharpest
case), #7 `Viewport` aligning WITHOUT taking focus, and #9 a focused
terminal panel not suppressing the document viewport.
**Every one of the nine pins was falsified by revert.** Two needed a
second attempt after the first bite came back green.
**P2-3 — stale docs.** `StatuslineEvaluationTarget::Semantic`'s
documentation described evaluating only the focused window; it now
describes the document-plus-side fan-out, the capture order, the
identity-selection requirement, and that `active` reports actual focus.
The ledger's Stage 2A entry is corrected to five commits, 2,014 CRDT
tests, and 16 acceptance tests.
Two clippy findings the refactor introduced were fixed:
`document_buffer_to_follow` is `crdt`-gated to match its only caller,
and the `CursorByte` guard collapses into one `if`.
Gates: fmt clean; workspace clippy clean; 1,832 default + 2,014 CRDT
library; Stage 2A 16; Stage 1 46; statusline 8; m11_5 2; GPU initial
target 14; terminal config 12; folding Stage 2 48; vterm 1/2 10 / 6;
M4 121; required GPU 202; `git diff --check` clean.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Stage 2 of docs/terminal-config-and-copy-mode-framing.md (rev 4,
approved). `M-x terminal.copy-mode`, or `C-t` in a terminal buffer —
physically `C-c C-t`, since every unescaped key goes to the child —
materializes the retained scrollback into an ordinary read-only,
path-less buffer, with `g` to re-snapshot and `q` to return.
No protocol change.
Materializing is the whole design. isearch, motion, selection and the
kill ring work with no new substrate because the snapshot is a rope, so
SearchStore and the existing match painting apply unchanged. And "keys
must not reach the child" dissolves structurally rather than being
guarded: the transport arm keys on is_terminal(buffer_id), and a
snapshot is not a terminal, so the arm never fires. The
dispatch-shadow count stays at six and describe-key keeps telling the
truth — asserted directly, since that is the observable difference
between the buffer-local idiom and a shadow.
One serializer, not two (Q#TC7). `copy_retained` builds a whole-range
selection and hands it to `copy_selection_bytes`; a second walk would
re-derive soft-wrap joining, wide-glyph continuation, cluster bytes and
per-row trailing-blank trimming, and the two would drift. Four unit
pins in view.rs assert exact bytes against the same projection fixtures
that pin the serializer itself.
Q#TC6a is implemented as two calls, and the second is the load-bearing
one: an intercept guards dispatch only, and no Lua binding sets
Buffer::read_only, so set_round_trip_input is what keeps a replica
frontend from applying optimistically and emitting an op that would
pass ensure_writable and mutate both sides. Acceptance 16 pins that
UNGATED, because CI never compiles the crdt feature.
Eight of nine criteria. Criterion 17's semantic-frontend end-to-end pin
is deliberately absent: the optimistic apply lives only in
pmacs-gpu/src/main.rs and the headless SemanticClient has no optimistic
path, so a faithful test needs the real GPU binary — the a37
foundation, which CI never compiles, silently returns ok when the
binary is unbuilt, and is load-sensitive. Both halves of the mechanism
are pinned ungated instead (16, and 16b for the hazard); the wire-level
half stays an explicit obligation of the CI crdt-coverage lane.
Substrate fact found while wiring lifecycle: TerminalManager::prune
REACTS to a buffer already gone from the registry rather than removing
one, so a child exiting leaves both the terminal and its snapshot
alive. That is why on_removed is a sound teardown hook, and why a
finished command's output stays readable.
Five bites, five different wrong implementations, each failing exactly
one test: removing set_round_trip_input fails acceptance 16 in the
DEFAULT configuration; a naive independent serializer fails all four
unit pins, with the diffs naming each drift mode; making re-invoke
create a fresh buffer fails 18; dropping the kill-with-terminal
teardown fails 18; removing the intercept fails 16b.
COHERENCE.md: §6 gains this as the worked example that a modal-looking
feature need not become a shadow; §11 records the scope="global"
deferral's second live case, making the argument for both registry
deferrals cumulative; §2 step 8 gains copy mode and keeps the
still-missing close command named.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016gGQC6eqHJVbZJ5Hg7aLer
Integrates canonical `main` @ `cf54270` and closes every finding.
**P1-1 — a stale document `Pointer` stole focus from the panel.** Real
bug. `align_primary_document_window`'s unknown-buffer arm returned
`Some(window)` despite aligning nothing, so #8's activation focused the
document *before* `dispatch_pointer` rejected the mismatched buffer. It
now returns `None`: alignment did not happen, so no caller may treat it
as a document gesture.
Pinned through `handle_dispatcher_event` — the real dispatcher seam —
because the defect lived in the PAIR of alignment and activation, not
in either alone. **The first version of that test was vacuous**: an
unregistered session is dropped at `daemon.rs:1962` (#148's
membership check) before the aligner runs, so it passed with the bug
restored. It now registers a real semantic session and fails with
exactly the reported symptom, focus moving `WindowId(2)` →
`WindowId(3)`.
**P1-2 — the approved A2A-2 fan-out was missing.** The semantic target
returned one context. It now captures the primary document PLUS the
visible side window, each provider invoked once, with a
derived-hidden side omitted (Q#BP2b — no mode line to paint, so no
callback should run for it). The acceptance asserts `windows.len() == 2`.
This exposed a second defect the finding did not name: the consumer
selected segments with `.find(|w| w.context.frontend_id == frontend_id)`
— the FIRST context for the frontend. With two contexts that silently
depended on capture order and could have shipped the panel's mode-line
text as the document status band. `emit_statusline_segments` now takes
the document `WindowId` and selects on window identity.
**P1-3 — the census suite tested the authority, not the consumers.**
Confirmed: reverting a producer to `active_window_for` left all ten
tests green. Added consumer-level pins that drive the real producers
through `SemanticRenderState::render_frame` with a panel focused, plus
the terminal-declaration guard. Bite-verified: reverting the
`LineNumbers` routing now fails
`consumer_line_numbers_follow_the_document_not_the_focused_panel`.
**P1-4 — main integrated.** The textual conflict was `docs/active-work.md`
(both lanes rewrote the same region; the terminal-config lane is kept
whole and the bottom-panel heading updated). `src/editor.rs` auto-merged,
and the full gate suite was rerun on the merge result.
**P2-5 — the painter test was vacuous.** A fixed-point check that
survived deleting `window.text_view.render`. It now asserts each of the
four extracted outputs actually appears: buffer TEXT, the line-number
GUTTER (with line numbers explicitly enabled, rather than dropping the
assertion), the window MODE LINE, and a returned caret. Bite-verified
by deleting the render call.
**P2-6 — the stale fold-projection claim is corrected.**
`src/window.rs`'s `fold_projection` doc no longer asserts that a
semantic session never enters `paint_frame`; it records that the panel
band breaks that premise and that the extracted painters take the map
as a parameter.
Gates on the merge result: fmt clean; workspace clippy clean; 1,832
default + 2,010 CRDT library tests; Stage 2A acceptance 13; Stage 1 46;
statusline 8; m11_5 2; GPU initial target 14; terminal config 12;
folding Stage 2 48; vterm 1/2 10 / 6; M4 121; required GPU 202;
`git diff --check` clean. `vterm_stage3_acceptance::a37` remains the
pre-existing flake measured on the base commit.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
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.
Bottom-panel Stage 2A, second half (Q#BP8, Q#BP17). Still no protocol
change and no behavior change: `paint_frame` builds the same fold map
it always did and passes it in, so grid rendering is unchanged.
Two extractions, both taking the fold map as a **parameter** rather
than building it:
- `prepare_window_cursor_visible` — the active-window auto-scroll
clamp. The panel band (2B) runs this for its own window when that
window owns focus, and leaves a passive panel's `view_top` alone.
- `paint_window_content` — the per-window document body: text, gutter,
overlays, selection, and the mode line. The panel paints into a
panel-sized grid at the same origin-agnostic `Viewport`, so this is
that body lifted out, not a second painter (Bet B2').
The parameter is the point (Q#BP17). Folding built its per-window map
ungated on the premise that "a semantic session never enters
`paint_frame`", which the panel band breaks. The panel path must pass
`None` for a frontend whose `fold_projection` is false, and must not
call `EditorCore::fold_map_for_window` — that gates on the **active**
frontend, which is right for command-time reckoning and wrong for
painting another frontend's panel.
`tests/bottom_panel_stage2a_acceptance.rs` — 10 tests. The negative
half is the load-bearing half, so Projection assertions are paired
with focus-class assertions taken in the SAME state:
- `focus_and_projection_disagree_in_the_same_state` is the key one:
with a panel focused, the focus authority must name the panel while
the projection authority names the document. Routing the focus class
through `primary_document_window` fails this even though every
Projection test still passes.
- The statusline pair pins the split: the LOOKUP resolves the document
window while `active` reports actual focus, with a non-vacuity twin
that flips `active` back to true when focus returns.
- The extraction pair pins cells, the returned cursor, the focused
window's `view_top`, AND a passive window's untouched scroll —
identical cells alone would not catch a clamp that moved to the
wrong window on a single-window frame.
- `the_panel_fixture_really_builds_a_side_window` pins the fixture's
own precondition, since every other test is worthless if
`focused_panel` silently produced an ordinary split.
One crdt-gated caller of the old `align_semantic_window_to_buffer` was
updated; it compiles only under `--features crdt`, which is the
config CI never runs.
1,832 default + 2,009 CRDT library tests, 10 new acceptance; fmt and
workspace clippy clean.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Five findings, all real. The blocker and both majors are the same
mistake in three places: a claim asserted somewhere cheaper than where
it actually lives.
COHERENCE.md was stale in four places, not the three reported. Step 8
still read "no keybinding" and §11 still read "five settings", but §6's
dispatch table also still cited `is_terminal_escape_chord` — a symbol
this branch deletes. §25 requires that update to ride the PR, so a PR
changing audited ground truth has to re-grep the audit for its own
symbols, not only for its topic.
Acceptance 5 asserted a registry round-trip, which is a test of the
registry: it stayed green with the setting's only consumer deleted. It
now opens a real terminal whose child overflows the 24-row screen,
scrolls the view to its oldest retained row, and asserts LINE001 is
present at 10,000 and absent at 0.
Acceptance 8a waited for the session count to fall, which the rejected
editor-side cache map satisfies exactly — a map with no purge hook
leaks while sessions drain. Adds `TerminalManager::escape_caches()`, the
lifetime half of Q#TC4c's contract that `escape_parses` cannot cover.
`table.sort` over `pmacs.terminal.profiles` raised "attempt to compare
number with string" on the unknown-profile path whenever the user's
table held both a string and a numeric key, replacing the exact
diagnostic being asked for; `%q` raised likewise on a non-string
`profile` argument. Both are partial functions applied to user input on
a diagnostic path.
Also corrects the framing's status line, and a status message whose
embedded whitespace run had survived a rustfmt reflow.
Three new bites, each falsified by revert: deleting the scrollback
consumer fails acc5 and only acc5; restoring the raw-key sort
reproduces the comparison error verbatim; and implementing the rejected
map fails the new acc8a at left: 2, right: 1 while passing the old
session-count version.
Merges githubsucks/main @ ccf29e3.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016gGQC6eqHJVbZJ5Hg7aLer
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 4 review: one P1, and it is the same defect for the FOURTH time.
`pmacs.lsp.config.lean4` is a single global entry, so swapping its
command invalidates **every** Lean buffer and **every** Lean server —
Q#LN15 gives one server per project root, so there can be several.
Rounds 1-3 each repaired one buffer and retired one server, and round 3
shipped "repair the armed target, strand the rest": status and config
said fallback while a second open Lean buffer stayed on the retired
command, and a second project root's server stayed live.
The shape that actually holds:
* **Retire ALL `lean4` servers on latch**, not the one the probe
happened to name. `probe.primary` identifies the server the VERDICT
is about; it was never the set of servers the swap invalidates.
* **Repair each buffer lazily and at most once**, when it becomes
active — on `buffer.after-switch` and on the tick. `_attach_buffer`
is an active-buffer-only seam, so a global swap cannot be applied to
every open buffer at once; it has to be applied as they surface.
lsp.lua's own `after-switch` re-pushes views but does not rebuild a
stale attachment, so nothing else covered this.
* The **once-per-buffer bound** is load-bearing: without it a fallback
that also fails to spawn would retry every tick forever — the
round-2 defect, which a naive global repair loop would reintroduce
for every buffer instead of just one.
* `shutting-down` is deliberately not treated as stale. It is still
live by `server_is_live`'s reckoning, so attaching would early-return
the stale record and burn that buffer's single attempt on a no-op.
P2: argument-inclusive attribution was implemented in round 3 but pinned
only by "contains the command name", so a mutation dropping every
argument passed. Now asserted against the exact `<command> <args>`
string.
Also fixed a vacuous assertion this refactor created: a test checked
`_probe.reattach_from == nil` for a field that no longer exists, which
reads as nil and passes for nothing. It now asserts a positive count of
recorded repair attempts.
Three bites, each against 73587b0: repair only the armed buffer -> the
second buffer stays on `lake`; retire only the named server -> one live
stale server remains; drop arguments from attribution -> the exact-string
assertion fails.
The ledger records a second durable lesson beside the vacuity one: **a
scope error repeats until the scope is named.** Four rounds of locally
correct fixes, none of which asked what the config swap invalidates.
When a change edits shared state, enumerate everything derived from it
before repairing anything.
Round 3 review: two P1 asynchronous-correlation defects, with the
focused suite at 25/25 while both were live.
**1. A late version verdict retired nothing and claimed success.**
`probe.watching` is cleared the moment the server initializes — it is
failure-polling state. A slow `lake --version` landing after a
successful initialize therefore reached `fire_latch(nil)`, which retires
nothing: `_attach_buffer` found the still-live primary attachment,
early-returned it, and the retry counted that as done. Status said
"falling back", the config named the fallback, and the buffer stayed on
the old server.
**That is the round-1 silent no-op arriving through a third event
ordering** — first as "no re-attach at all", then as "re-attach cleared
by an unrelated buffer", now as "re-attach satisfied by the server we
were supposed to replace". The fix separates the two facts that were
being carried by one field: `probe.primary` is the server the verdict
applies to and survives initialization; `probe.watching` is the
failure poll and is cleared by it.
The existing fixture could not reach this ordering at all — its `serve`
sleeps, so the primary can never initialize before `--version` returns.
The new one execs the fake LSP for `serve` and delays 0.6s before
reporting 3.0.0.
**2. `buf_key` was the most recently loaded Lean buffer.** Written on
every Lean `buffer.after-load`, so a second Lean file opened before the
verdict became the rebuild target while the latch still watched the
FIRST buffer's server. Target buffer and primary server are one fact and
are now armed together, exactly once. Both files in the new test share a
package, so mis-targeting shows up as a stranded buffer rather than as
two unrelated servers.
**3. The failure message hardcoded `lake serve`** after the latch became
command-agnostic, telling a user whose `my-lean-wrapper` failed to go
debug lake. `configured_command()` names what is actually configured,
arguments included.
**4. The ledger** now records all fifteen bites across the three rounds,
both prior review rounds' findings (the round-2 block was lost when an
earlier edit script aborted before writing), and the durable lesson.
That lesson, recorded for the handoff: **six tests across three rounds
were written, ran green, and pinned nothing** — caught only by biting.
The shapes are enumerated in the ledger; the rule is that a test is not
evidence until the mutation it targets has been shown to fail it. Two
of the six are subtle enough to be worth naming here: a bite that
RAISES is swallowed by the hook's pcall and "passes" for the wrong
reason, and a fixture whose `serve` sleeps cannot reach any ordering
where the primary comes up first.
Stage 1 of the terminal config/copy-mode arc. The terminal had no
configuration surface at all: the command hardcoded $SHELL, scrollback
was a per-open argument only, and the escape chord was a literal in
Rust. No protocol change.
Profiles are a raw Lua table, pmacs.terminal.profiles, not a registry
setting: ConfigValue is four scalars with no table kind, so profiles
join pmacs.lsp.config and pmacs.pair.sets until table-valued settings
exist. The registry gains three scalars whose defaults reproduce the
previous behavior exactly.
Field resolution is explicit open argument, then profile field, then
scalar setting, then $SHELL. env MERGES, with explicit entries
overriding the profile's, because first-wins there would silently drop
half a user's environment. An explicitly named profile that does not
exist is an error even when terminal.default-profile is valid, so a typo
cannot silently fall back.
The two open-time settings resolve through the GLOBAL chain, because
they are read before the identity buffer exists and no caller could have
pinned a local override on a buffer that does not yet exist. Only
terminal.escape-key resolves per buffer, which makes a per-terminal
escape a supported feature.
The escape key is parsed at most once per (terminal, config epoch), and
the cache lives on TerminalSession so its lifetime is the terminal's,
with no purge hook to forget. The epoch alone is not a sufficient key:
it does not advance when focus moves between two terminals with
different buffer-local values, so an epoch-only cache serves one
terminal's chord to the other. An unparseable value falls back to C-c
and reports once per terminal per effective invalid value through the
status line, because a terminal with no escape chord cannot be escaped
to fix the setting that broke it.
Repeating the escape now sends THAT chord to the child through the
ordinary key encoder, rather than a hardcoded ETX. With an escape of
C-x, the previous code sent Ctrl-C and made literal Ctrl-X unreachable.
C-c t opens a terminal. COHERENCE Priority 1 names a terminal
keybinding, and section 2 step 8 grades the terminal works-but-
undiscoverable; C-c is already a live global prefix, so this is a new
leaf rather than a shadow. It is unreachable from inside a terminal,
where C-c is the escape.
Acceptance is tests/terminal_config_acceptance.rs, deliberately NOT
crdt-gated so CI actually runs it. Four bites, each against a different
plausible wrong implementation: a hardcoded ETX fails acc6/9; an
epoch-only cache key fails acc7; a single last-entry cache fails acc8's
parse count; removing the invalid-value fallback fails acc10.
Two test-instrument notes worth keeping. cat -v is the echo probe
because the screen rejects C0 controls before they reach cells, so a raw
echoed Ctrl-X would be invisible. And the probe counts occurrences
rather than testing presence, because a single-character probe collides
with the child's own banner text.
Three P1 lifecycle defects and two P2s. The focused suite was 20/20 with
every one of them live, which is the part worth keeping.
**1. The crashed primary respawned forever underneath the fallback.**
Round 2 skipped the retire call for terminal servers to avoid corrupting
them — but the crash had already armed `next_restart_at`, and
`maybe_restart` fires on every elapsed backoff with no attempt ceiling.
The broken command kept respawning under the live fallback.
The right call depends on the state, and each is wrong for the other:
`forget` REQUIRES a terminal state and removes the client outright,
which also drops the restart timer; `stop` is for a live one and
corrupts a terminal one (its not-initialized branch parks it in
`ShuttingDown` forever). `retire_server` now dispatches on state.
**2. Re-attachment targeted whatever buffer was active when the
asynchronous verdict landed.** `_attach_buffer` is an active-buffer-only
seam, and "some attachment now names a different server" is satisfied by
an unrelated Rust buffer — clearing the retry and leaving the Lean buffer
stale forever. The initiating buffer is now captured and the retry waits
for it.
**3. A failing fallback retried every tick forever, silently**,
contradicting acceptance 27's promise that a second failure surfaces.
"Waiting for the old server to go" and "attempting the replacement" are
now separate: once the old one is terminal or gone, the replacement is
attempted EXACTLY once, and a spawn failure is reported.
**4. The Lake version parser was being applied to arbitrary wrappers.**
`version_below_3_1` encodes lake's output contract; a working
`my-lean-wrapper` reporting "wrapper 1.0" would have been replaced
despite its server initializing fine. The version probe is now gated on
the command's basename being `lake`. The FAILURE latch stays
command-agnostic — that one keys on the server actually not starting,
which is true of any command.
**5. An unconfigured Lean server was reported as a failure** and latched,
poisoning the session so a later configuration could never take effect.
Absent config or command now means disabled; only a configured command
that produced no attachment is a failure.
**6. The ledger recorded pre-fix counts** after the fixes were pushed.
Now 25/25 and 3,214. That is the #161 fmt-blocker error in a slower
form: verification must describe the pushed tree.
Sign-offs requested in review: `M.fallback` is now `M._fallback`, an
underscored test seam, and its idempotence check compares args as well as
command — the same command with different arguments is not "already
applied". Dropping the `command ~= "lake"` guard stands for the failure
latch only.
Five regression tests added, and **three of them were too weak on first
write; only bite-testing found it**:
* asserting "no live non-fallback server" misses a respawn loop,
because a respawning server sits in `crashed` most of the time —
`attempt` is the observable that counts respawns;
* returning to a buffer with `find_or_open` re-fires
`buffer.after-load`, which repairs the attachment regardless of the
code under test — `switch_buffer` is the honest return;
* a MISSING command fails synchronously inside `after-load` where the
rebuild happens inline, so the async race cannot occur — only the
probe path exercises it.
Each of the five now fails against the exact round-2 mutation it targets.
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.
Framing Q#LN7, Q#LN8, Q#LN16; acceptance 22–28, 24a/24b, 35, 36, 36a, 37.
Stacked on Stage 3a (#167), whose notification/response seams and
`pmacs.fs.canonicalize` this consumes. No protocol change; the only Rust
outside the test helper is one `include_str!` line.
**The Lake-aware root (Q#LN8).** `pmacs.project.detect` cannot express
this rule — it is innermost-wins by construction, and a Lake package's
`lean-toolchain` sits at the outermost level, so a file under
`<pkg>/.lake/packages/dep/` belongs to `<pkg>`'s server rather than
`dep`'s. The resolver walks up collecting markers and returns the
outermost, stopping at `pmacs.project.search_boundary()` so a stray
marker above a fixture cannot leak in.
Two things about the marker test are easy to get wrong in opposite
directions, and both are pinned. `io.open` **succeeds on a directory**,
so a truthiness check accepts a `lean-toolchain` directory; but
requiring a non-nil read rejects an **empty** `lean-toolchain`, which is
a legitimate marker — `locate-dominating-file` semantics are existence,
not content. The discriminator is `read`'s second return: decline only
on a non-nil error. Acceptance 24a and 24b each fail against the
implementation that satisfies only the other; both bites are recorded.
The root is canonicalized once up front, because a configured root
reaches `file_uri_for` verbatim and that URI is the affinity key (#161).
Canonicalizing the starting directory suffices — every ancestor of a
canonical path is canonical, since the walk only strips components.
**`lake serve` with a lazy probe and a one-shot latch (Q#LN7).** Nothing
runs at init: `pmacs.lsp.config` is declarative, and spawning a process
at startup for every user, Lean-using or not, is the cost rev 1 refused.
Both the probe and the server spawn are gated on a real Lean attachment.
The probe cannot gate the first attach — there is no blocking process
run, so its verdict arrives after `ensure_server` has already decided.
Hence the optimistic spawn, with the probe and latch correcting it. A
non-zero probe exit is deliberately NOT a trigger: §2.9's elan-shim case
makes `lake --version` fail on machines where `lake serve` still works,
and the server-failure latch covers that better. The probe answers only
the question failure detection would answer slowly — an old-but-working
lake that starts a useless server.
The latch stops the failing server **before** spawning the fallback, and
that ordering is load-bearing rather than defensive: the spec default is
`OnCrash`, the termination handler never consults the exit code, and
`maybe_restart` has no attempt ceiling, so a broken `lake` respawns
forever underneath the latch. `pmacs.lsp.stop` sets `restart = Never`,
which is what disarms it. Bitten: removing the stop fails acceptance 36.
The swap rewrites `command` and `args` only, so a user's `env`,
`settings`, `init_options` and `root` survive — a wholesale table
replacement would discard their `init.lua` at the moment they are least
likely to notice.
**`waitForDiagnostics` (Q#LN16)** resolves through Stage 3a's response
seam, with `M-x lean.wait-for-diagnostics` on top. `$/lean/fileProgress`
subscribes on the notification seam and is pinned end-to-end through a
new `leanprogress` mode on the fake server rather than by calling the
handler directly — the wiring is the only part that can break.
**Attribution (COHERENCE §9/§1.2).** The probe spawns as
`lean:lake-version-probe`, so a user wondering why their editor touched
`lake` finds an owner in `pmacs.process.list`. The latch reports through
`pmacs.editor.set_status` — the channel that exists — and acceptance 36a
observes that channel, so a report made only through the undefined
`pmacs.error` would fail it.
**Stage 1's acceptance 12 is updated, half superseded.** It asserted
`pmacs.lsp.config.lean4 == nil` to guard against a Stage-3 front-run;
Stage 3b is that stage, so keeping it would pin the opposite of the
intended behavior. The half that survives is the one about restraint,
and it matters more now: constructing an editor spawns nothing even
though the config exists and names `lake`, and opening a Lean buffer
with no server configured spawns nothing either. That is what holds
Q#LN7's "not at init" promise.
Bites recorded, all against the committed tree: bare `io.open` -> 24a
fails, 24b passes; require-non-nil-read -> 24b fails, 24a passes; no
canonicalization -> the symlink case spawns two servers; no stop before
fallback -> acceptance 36 fails.
CI round 1: both macOS jobs failed on the acceptance case added last
commit. APFS enforces valid UTF-8 in filenames, so `std::fs::write` with
a 0xFF byte in the name fails with EILSEQ ("Illegal byte sequence")
before `pmacs.fs.canonicalize` is ever called. The fixture cannot be
built there.
That is a filesystem refusing to represent the case, not a behavioral
difference: the subject — `to_str()` returning None for a non-UTF-8
resolution — is platform-independent Rust, and the Linux run pins it.
`#[cfg(unix)]` was the wrong granularity; review had asked for unix
gating on the symlink tests and I applied the same gate here without
checking whether the filesystem, rather than the API, was the
constraint.
Gated `#[cfg(target_os = "linux")]` with the reason in place, rather
than skipped at runtime, so a future failure here is a real failure and
not a silent no-op.
Ledger records both CI-round facts: this one, and that
`composition_overhead_under_ten_percent` is load-sensitive under a
parallel workspace sweep (it reported -4.6% realistic overhead in the
same run that tripped its 10% budget at 18.8%, which is noise, not work).
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.
Main moved again while this lane was gating: the GPU terminal-input fix
merged as #166. One conflict, in COHERENCE.md's journey table, resolved
as the union -- this lane owns step 7's file half, #166 owns step 8's
GPU-terminal addendum.
Multi-root LSP affinity merged as #161 (`main` @ `46a1b8f`) while this
lane was in review, which made the PR conflict -- and a conflicting PR
has no merge ref, so GitHub silently stopped running CI on it after the
first push. Integrating rather than rebasing, per the #135/#137
precedent: the review anchors stay addressable and every gate is rerun
against the merged tree.
One conflict, in COHERENCE.md's in-flight list, resolved as the union of
both truths -- and #161 is now merged, which its own text still called a
PR.
The overlap to watch is `src/lua_bindings/mod.rs`: #161 widened the
`lsp.list()` row builder while this lane added `pmacs.path` and the
read_dir listing conversion. The merge was textually clean, which the
folding arc's lesson says is not the same as compiling, so the full gate
suite reruns from here.
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.
The dispatcher loop applied BOTH terminal-layout syncs to EVERY attached
frontend. A semantic session satisfies both conditions, because it has a
term_sizes entry from AttachRequest and a semantic terminal declaration,
so its PTY was resized twice on every tick forever: the grid arm
installed the TUI placement size, the semantic arm installed the declared
content rectangle, and each arm's own idempotence guard only ever saw the
size the other had just written. The child took a SIGWINCH storm at tick
cadence and the screen reflowed continuously, which is what made typing
into a GPU terminal impossible while output kept flowing.
The grid arm is also the only per-tick controller-liveness release a
semantic frontend gets, so simply skipping it for those frontends trades
one defect for another: a GPU window that switches away from its terminal
would hold the controller forever, and no peer could resize that PTY
again. The semantic arm cannot take over that job, because the
buffer-follow snapshot clears the viewport declaration that would drive
it.
sync_terminal_layout is therefore split into a frontend-kind-neutral half
(panel reconciliation plus controller liveness, which read only views,
windows and the controller) and a grid-only geometry half (TUI placement
plus the resize). The dispatcher runs the neutral half for every attached
frontend once per tick, then exactly one geometry arm per frontend kind.
sync_terminal_layout survives as the composition of both halves, so the
in-process editor loop and LOCAL are unchanged.
The loop body is extracted into sync_terminal_layouts_for_tick, which
makes the grid/semantic exclusivity structural rather than two adjacent
ifs, and lets the tests drive the real loop body instead of a
re-implementation.
The release that fires when a window has no placement stays in the grid
half deliberately: a semantic frontend has no window_placements entry at
all, so moving it into the neutral half would release a GPU session's
controller on every tick.
Bite-verified against two pre-images, because one is not enough here --
the naive guard fixes the storm and introduces the controller leak, so a
single revert would score the fix complete when it is not:
pin main naive guard split
settle (acc 2+3) FAIL pass pass
controller release (acc 6) pass FAIL pass
grid still resizes (acc 5) pass pass pass
Real-path acceptance: a quiet child that counts SIGWINCH reports 144
frames in 4 s and WINCH 1..12 on screen against the pre-fix tree, versus
a settled screen with the fix. Acceptance 4 (input reaches the child and
returns) is a keep-working pin and passes on both sides -- key transport
was never the defect.
No protocol change; stays v20.
F1 (real, small-window misbehavior). `dired.revert`'s re-seat runs after
the read settles, and `pmacs.editor.move_to_line` is AMBIENT -- it moves
whatever window is active. A user who switched buffers (or hit `q`)
while the re-read was in flight had an unrelated buffer's cursor moved
to a line index that only means something in the dired listing. The
paint was already safe because it names its buffer; the seat now runs
only while dired is still the active buffer, and `seat_cursor`'s doc
says which callers are unconditionally in the right place and why.
Pinned by a test that starts the revert, switches to a six-line file
before the pump, and asserts that buffer's cursor never moved -- and
that the dired buffer still reverts when it IS active.
F2 (a trap set for Stage 3). `fmt_size` used `%10d`, so a size past ten
digits -- 10 GB and up, ordinary for VM images and core dumps -- widened
the field and shifted mtime and name right on that line alone. Cosmetic
today, but `_layout` is exported as a contract and Stage 3's
column-classifying intercept is planned against it. It now takes
`fmt_mtime`'s discipline: exact bytes while they fit, else a
fixed-width magnitude, so precision yields to the invariant rather than
the other way round. This is not the deferred human-readable column --
the exact count still shows right up to where it cannot. Pinned with a
sparse 12 GB fixture that skips if the filesystem refuses it.
F3 (honesty and a doubled read). The symlink arm claimed the probe cost
"one syscall"; it was a full `read_dir` -- opendir plus one lstat per
child -- and on success `open_directory` immediately read the same
directory again. Since `open_directory` reads before touching editor
state and raises having changed nothing (acceptance 15's invariant), its
failure IS the "not a directory" answer: the probe is gone, one read
remains, and the comment says what it actually does. New test pins both
arms -- a symlink to a directory descends under the path the user walked
(canonicalization is lexical, so the link is not resolved), and a
symlink to a file opens with the target's contents.
F4 (deliberate failure mode). A tolerant listing recorded readdir
iterator errors without bound, and `std::fs::ReadDir` need not terminate
after yielding one. Cancellation is NOT an adequate backstop here --
which is the reason for a constant rather than a comment saying it is: a
dired listing carries no supersede key, so nothing cancels it. A
directory whose iterator produces nothing but errors now fails with the
last error the way an unopenable directory does, after
READDIR_MAX_CONSECUTIVE_ENTRY_ERRORS; the counter resets on any entry
that materializes. Documented as untested and why: faking a failing
iterator needs the walk generic over it, a refactor with no other
consumer.
Smaller notes, all taken: READ_ONLY_LIMIT renamed NAME_VARIANT_LIMIT (it
caps `<2>`..`<99>`, nothing read-only); `fmt_perms`' omission of
setuid/setgid/sticky documented as a decision tied to the M8.3 fixture's
nine-bit parser; `format_outcome` binds the slice in the pattern instead
of re-traversing; and `pmacs.path.canonicalize`'s `to_string_lossy` is
noted as inside the existing non-UTF-8-path deferral rather than an
exception to it.
Process note, learned the hard way twice now: the round-1 dired.lua
fixes were briefly wiped because a mutation-bite helper restores with
`git checkout --`, which reverts to HEAD -- so a fix must be committed
BEFORE it is bitten, not after.
`describe_key_identifies_every_default_binding` iterated every binding
in the stack and asserted `pmacs.describe.key` resolves it context-free.
That held only because no builtin had ever bound a mode-scoped key:
dired is #129's first non-detection consumer, so its `n` / `p` / `g`
correctly resolved to nothing and the test went red on the feature
rather than on a defect.
It now sets the effective context per binding -- the mode for a
mode-scoped default, and explicitly NO mode for a global one, because a
leaked mode legitimately shadows a global chord of the same name
(dired's `RET` shadows `edit.newline-and-indent`, which is the point of
the mode) and would make the assertion compare the wrong pair. A floor
assertion keeps the new arm from going vacuous if the last mode-scoped
default is ever removed.
Also corrects a doc comment rather than leaving it to be believed:
acceptance 3c does not pin the descent ROUTING. Dired holds focus in its
own panel, so a raw `switch_buffer` lands in the same window and the
mutation is vacuous against that test; dedication is what distinguishes
the two paths, so the discriminating pin is the dedicated-panel test
next to it. Verified by mutation, not assumed.
Dired is the file surface, not a rider on one: before Stage 0 (#162)
pmacs had no way to open a file by path, and browsing is the half a
user reaches for when they do not already know the path. Stage 1 ships
the view.
builtin/runtime/dired.lua: one buffer per directory named by the
canonical path (Q#DR2) with an ownership check before any paint (F7);
read-only intercept plus round-trip input (Q#DR3); a `dired` major mode
carrying mode-scoped keys (Q#DR8) -- RET/f visit, ^ parent, n/p, g
revert, q quit, s sort; cursor re-seated by basename across every
wholesale repaint (Q#DR9); file visits through
`pmacs.window.display_file` and directory descent through dired's own
window (Q#DR10); `C-x d` / `C-x C-j`; and `dired.kill-when-opening`
through the config registry.
Two Rust changes, both narrow:
* `read_dir` grows per-entry tolerance behind an opt (Q#DR6). Five
per-entry conditions used to fail the entire listing, so a plain
refresh of a busy directory could just fail; the module doc's claim
that a tolerant wrapper was "the package's job" was false, because
the primitive hands Lua one structured error and no partial vec.
Per-entry readdir/lstat/readlink failures and non-UTF-8 symlink
targets now land in an `errors` channel; parent-level failures and
non-UTF-8 *names* stay fatal. The tolerance travels in the settled
payload, so the Lua boundary keeps the bare-array shape the frozen
M8.2 fixture consumes and never has to look the job back up. The read
ops' opts parsing now rejects unknown keys, so a typo'd `tolerant`
cannot silently degrade to the fatal contract.
* `normalize_buffer_path` is exposed as `pmacs.path.canonicalize`
rather than mirrored in Lua. Q#DR2 named exposure the preferred end
state; it needs no borrow plumbing, so dired's name-dedup and
`display_file`'s `find_buffer_for_path` dedup cannot fork, and the
mirror's Stage 2 removal is not owed.
tests/dired_acceptance.rs covers framing items 1-16 (22 tests), driven
through real key dispatch. Item 17 is the m8_1/m8_2/m8_3 gate.
One framing claim is corrected by the substrate: R2-3 expected a
dedicated dired panel to carry its dedication across a descent, but
`display_buffer` never replaces the buffer in a slot dedicated to
another one -- it discards every side-specific parameter and falls back
to the document window (Q#BP3 2.iii). Dired does not try to unpin the
user's panel; both arms are pinned.
Levi Neuwirth