`main` moved through #158-#166 (Lean 4 Stage 2, COHERENCE.md, find-file,
the dired framing and Stage 1, the GPU terminal input fix) while this
documentation branch waited. Both required docs conflicted; neither
conflict was a code signal.
Resolution:
- `docs/active-work.md`: main's ledger is the base — every lane it has
gained since this branch was cut is kept verbatim. Only the
bottom-panel lane is replaced with this branch's "Stage 1 MERGED;
Stage 2 (GPU band) is next" section, and only the bottom-panel entry
is added to "Closed since the last snapshot".
- `docs/agent-handoff.md`: main's version is the base. This branch's §1
bottom-panel bullet, its §1 roadmap Arc 7 entry (which also records
that DAP is now unblocked), and its four §5 ops lessons are inserted
at their anchors.
One repair rides along. Main's `docs/agent-handoff.md` carried a
garbled fragment at §1: a duplicated, truncated "GPU initial target
LANDED — #148" bullet whose body was the tail of the old head-of-`main`
anchor bullet, leaving the `SUPPORTED=[6..=20]` protocol enumeration
orphaned mid-sentence. The fragment is removed and the enumeration is
restored as its own bullet.
No code changes; the merged tree's non-doc content is main's.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
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.
#165's own commits could not update the handoff snapshot to name the
merge that contains them, so the protocol obligation lands here.
- `docs/agent-handoff.md`: absorb the dired lane into §1, replacing the
placeholder that promised exactly this. The bullet carries Stage 1's
durable substrate facts — why the tolerant `read_dir` had to be Rust,
why exposing the core normalizer beat mirroring it in Lua, the
fixed-width `_layout` contract Stage 3 reads offsets from, the
ambient-action buffer guard, treating a failure as the answer instead
of probing, the per-entry error cap, the first mode-scoped keymap and
the pre-existing test it broke, and the dedication a descent does not
carry. Refresh the head-of-`main` anchor and the last-updated line.
- `docs/agent-handoff.md` §5: two ops lessons that cost real time. A fix
must be committed before it is bitten, because `scripts/bite` restores
by `git checkout --` and reverts to HEAD; a CONFLICTING PR runs no CI
at all, because `pull_request` workflows build a merge ref GitHub does
not create while the branch conflicts, and nothing reports the absence.
- `docs/active-work.md`: remove the merged lane per update-protocol rule
4 and summarize it under "Closed since the last snapshot", keeping the
two forward items Stage 2 needs (the rename rebind is first-match-only
over a raw path, and Q#DR5's seam is the main-thread drain). Refresh
the canonical base. Flag the two lane headers that still call a merged
PR "IN REVIEW" — #161 and #166 — rather than editing lanes another
thread owns.
- `COHERENCE.md`: #165 is no longer a PR. Per §25 the audited claims this
work changed were updated when it landed; this corrects their tense in
seven places and the two prose lines that still asserted dired was in
flight.
- `docs/dired-framing.md`: status line to MERGED, and state plainly that
Stages 2 and 3 each still need their own framing.
Docs only; no code, no gate-relevant change.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0126d2sikA6jZpFin3rtLCSK
Third merge of main into the lane, at b889873 (GPU terminal input #166).
Unlike the first two this one produced NO conflict -- and it is the case
that shows why a clean git merge-tree is not a reason to skip
integrating. #166 lands 41 lines in pmacs-gpu/src/main.rs, the same
heavily-rewritten file as the first integration; the two edits merged
silently only because they sit in different regions of it (#166 is
entirely in the headless probe, this lane rewrites the render path).
Merging the PR on that clean auto-merge would have shipped a combination
no gate had run.
Reconciliation, run against what #166 actually added rather than against
a pass/fail: it adds 3 library tests, 2 to vterm_stage3_acceptance, and
0 to pmacs-gpu. Predicted lib 1,826 -> 1,829, CRDT 2,003 -> 2,006, GPU
unchanged at 202; that is exactly what ran. Suite count 91 -> 92 is
#161's new lsp_multi_root_acceptance binary. All three sides' markers
verified live in the shared file.
Also records an ops trap that cost hours this session:
m4_5_basedpyright_initializes_and_negotiates_encoding does not time out,
it hangs forever, parking a --workspace sweep at 38 of 92 suites with a
live basedpyright langserver child. The per-suite M4 gate already skips
it; the workspace sweep needs the same flag.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HZjWMjwPXhPbt9upku9mCk
The lane opened in the previous commit was scoped to the Vterm Stage 3
acceptance. Measuring it properly shows the problem is much larger and
not vterm-specific.
Comparing cargo test --list under CI's exact flags against the same
flags plus crdt: 3,024 versus 3,288. 264 tests are dark in CI, and the
single worst line is the library itself at 177 -- cargo test --lib
--features crdt is a required local gate that CI has never run. Ten
suites run zero or one test, including gpu_initial_target (#148's entire
acceptance, 1 of 14), gpu_invocation (#141's, 1 of 14), and a37, the
Stage 3 real-daemon/real-PTY/real-wgpu path that #135 built precisely
because a decoded-message fixture would prove none of the three fit
together.
The lane now carries the per-target table, the verified flag combination
for the fix, a two-part fix shape (a crdt leg on the test job, plus the
GPU-requiring suites onto the existing gpu-render job that already has
lavapipe), and an explicit instruction to sort deliberate exclusions
from accidental ones first -- some of the 264 are perf suites that are
ignored by default and belong to their own jobs, while m10_10_perf has
no ignore attribute and no job naming it.
docs/vterm-framing.md gains an as-framed audit section. The arc is
structurally complete and every test named in the Stage 2 verification
map exists, but criterion 22's "without thrash" clause was never pinned
anywhere -- the word appears nowhere in src or tests -- and that clause
describes exactly the defect #166 fixed. Of the nine Stage 3 tests, only
three drive a real daemon, so the six that construct EditorState
directly could never see a dispatcher-loop defect; a31 passes on the
broken tree for that reason. Four of the nine, including a37 and Stage 3
review round 1's own presence regression guard, do not run in CI at all.
The section also records what was not audited: section 11's blanket
claim about deferral safety covers roughly twenty items and none were
spot-checked.
docs/gpu-terminal-input-framing.md scores bet B2 true now that the
reporter has confirmed typing works, and retracts Q#GT5. The bash fixture
behind it does not reproduce in real use and was almost certainly
measuring its own timing rather than a product behaviour; it is marked
retracted rather than deleted so nobody re-derives it from an earlier
revision.
docs/agent-handoff.md section 5 gains the lesson the confirmation cost:
a daemon-side fix is not deployed until the daemon is restarted from a
tree containing it, and rebuilding a binary does nothing to a running
process.
No code changes.
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 advanced twice inside one review round (#161, then #166), the
second landing while the first integration's sweep was still running.
The ledger now names both integrations, how each doc conflict was
resolved, and the verification numbers for the twice-merged tree -- plus
the lesson that a lane in review against a fast-moving main reruns its
gates per integration, not per push.
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.
The handoff snapshot and the active-work ledger both still described the
GPU terminal input work as in review. Per their own update protocols this
should have ridden #166; it did not, because the review that surfaced the
CI-coverage finding came after that PR was already green, and expanding
an approved PR to carry a new lane would have been the wrong trade.
docs/agent-handoff.md section 1 gains the #166 entry: the split into a
frontend-kind-neutral liveness half and a grid-only geometry half, the
extracted dispatcher loop body, the trap about the no-placement release
that reads like liveness and is not, and why the one-line guard was
rejected.
docs/active-work.md moves the lane to "Closed since the last snapshot"
and opens a new one: the Stage 3 real-path acceptance is dark in CI.
The workflow never enables the crdt feature, so every crdt-gated
acceptance test is not merely skipped but never compiled -- which covers
a37 (real daemon, real PTY, real wgpu) since #135 as well as the two
tests #166 added beside it. The fix is one step on the gpu-render job,
but it needs its own lane because it would run a37 under lavapipe for
the first time, and neither its timing budgets nor its wgpu path have
been exercised on that adapter. The lane also asks which other
crdt-gated suites are dark for the same reason.
Recorded alongside it: #166's three unit pins are not crdt-gated and do
run under CI's exact flags, including the controller-release pin whose
only job is catching the plausible wrong fix, so the regression
protection is live even though the real-daemon evidence is local-only.
No code changes.
The module doc said an uncaught raise inside a `pmacs.async` coroutine
"goes to *errors*, not the status line". #161's COHERENCE finding shows
that is wrong, and in the worse direction: `pmacs.error` is never
defined in production, so `step()`'s guarded report is dead and the raise
falls through to a bare `error()` inside `pmacs._async.tick()` -- whose
result `EditorState::tick_async` discards with `let _ =`. The failure
reaches nowhere at all, and dired would look like it silently did
nothing.
So the per-coroutine `pcall` plus `pmacs.editor.set_status` is
load-bearing, not tidy, and the doc now says which channel is dead, which
is live, and that the acceptance suite observes the live one -- the
corollary COHERENCE draws from that finding.
The ledger records the integration, the reruns on the merged tree, and
the ops lesson that cost three CI runs: a conflicting PR has no merge
ref, so GitHub creates no `pull_request` run and nothing reports the
absence.
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.
Rev 5 said acceptance 34's second edge was a killed buffer. Implementing
it showed that is false: the Rust core 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**, so lsp.lua never tears an attachment down and the drain keeps
reaching that server. The premise (the drain builds its sid list from
`attachments`) was right; the inference needed attachments to be removed
on kill, and nothing removes them.
The reachable leak has the same root cause by a different path.
`attach_buffer` drops a sid from `attachments` the moment
`server_is_live` reports false and rebuilds against a fresh server — so
`crashed` / `stopped` is the event *least* likely to be drained, and an
event-driven purge leaks in exactly the case it exists for. The purge
therefore polls `pmacs.lsp.list()`, which enumerates the manager
directly. Acceptance 34's second half now exercises a server in **no**
attachment, which is the shape that discriminates: bitten, an
event-driven purge fails it while the attached case still passes.
§0.1 finding 6, Q#LN9, and acceptance 34 all updated; the wrong wording
is left visible with its correction rather than quietly replaced, since
the mistake is the useful part.
Ledger gains the Stage 3a lane: branch, worktree, what ships, both
corrected claims, the `install_async` load-order trap, the recorded
bites, the one knowingly unpinned guard, and gate results.
Two corrections found by bite-testing the suite rather than by reading
it.
**Acceptance 32 was mis-named.** It claimed to pin "the one-shot is
removed BEFORE invocation". Biting that — moving the removal after the
`pcall` — still passes, because `pcall` catches the raise either way and
the removal runs regardless. The before/after ordering is unobservable
unless a handler re-enters the drain, and nothing does. What the test
actually pins is that removal is **unconditional**: the bite that gates
it on `if ok then` fails 2 != 1, because the surviving registration is
invoked a second time by the purge. Renamed and re-commented to say so.
The implementation still removes before invoking, which is the right
defensive order; it is simply not what the assertion proves.
**The purge's generation check is defensive and untested**, now labelled
in place instead of reading as covered. Reaching it needs a crash and
its restart to both land in a gap with no `_async.tick`; the backoff is
500ms, so any tick in that window sees `crashed` and the
absent-or-terminal test fires first. Every attempt to stage it
deterministically exercised the `crashed` path instead. It stays as
insurance for a stalled editor, and says that about itself.
Bites recorded, all against the committed tree:
- removal gated on a clean return -> acc32 fails (2 != 1).
- purge driven by a `crashed`/`stopped` event seen in the drain, the
design the framing originally implied -> the no-attachment case fails
("never called"), while the attached case still passes. That is the
discrimination acceptance 34's second half exists for.
- a resolver without `pmacs.fs.canonicalize` -> two servers, pinned as
34b's own falsification.
Arc 8 Stage 3a (framing Q#LN9, Q#LN20). No Lean content: this changes
the event drain every LSP language runs through, and is split from the
Lean server work for the reason Stage 2 was.
**The seams.** `handle_server_requests` handled five `request` methods
and `initialized`, dropping every `notification` and `response` on the
floor. Dropping responses made `pmacs.lsp.send_request` a write-only API
from Lua — the reply was drained and discarded, so nothing outside
Rust's typed stores could consume one. Two new arms route to
`pmacs.lsp.on_notification(method, fn)` (persistent, method-keyed) and
`pmacs.lsp.on_response(sid, request_id, fn)` (one-shot). Both extend the
existing loop rather than opening a second `events_take` caller, which
would steal events from it.
A one-shot is removed **before** invocation, so a raising handler cannot
be re-entered. Every subscriber is `pcall`ed and a raise reports through
`pmacs.editor.set_status` per COHERENCE §1.2 — not `pmacs.error`, which
is defined nowhere in production. The notification list's length is
captured before the walk so a subscriber registering another cannot
extend the list being iterated.
**The purge is driven off `pmacs.lsp.list()`, not off a death event.**
The framing said acceptance 34's second edge was a killed buffer. That
was wrong, and scouting the implementation is what caught it: pmacs
fires exactly five hooks (`buffer.after-edit`, `buffer.after-load`,
`buffer.after-switch`, `frontend.detached`, `process.after-tick`) and
there is no buffer-kill hook at all, so `lsp.lua` never tears an
attachment down and the drain keeps reaching that server. No leak there.
The real leak is a different path with the same root cause. The drain
builds its sid list from `attachments`, and `attach_buffer` drops a sid
from that table the moment `server_is_live` reports false — rebuilding
against a fresh server. So the `crashed` / `stopped` event that should
trigger the purge is precisely the one most likely to go undrained. A
purge wired to that event leaks exactly when it matters.
`pmacs.lsp.list()` enumerates the manager directly and is unaffected by
attachment bookkeeping, so the purge polls it after each drain: a sid
that is absent, terminal, or running a **new generation** settles its
pending one-shots with an error. The generation check uses the `attempt`
field, because a crash-then-restart reuses the sid — without it a
one-shot would sit waiting on a reply the dead generation owed.
**`pmacs.fs.canonicalize`** (Q#LN20) is the one synchronous function on
`pmacs.fs`, and synchronous is the point: its consumer is a
function-valued `config.root` called from `ensure_server` <-
`attach_buffer` <- `buffer.after-load`, where there is no coroutine and
`pmacs.fs.stat`'s awaitable handle is unusable. It is installed from
`install_async` rather than `install_project` purely for load order —
`make_workspace` runs after `fs.lua` is evaluated, so a canonicalizer
placed there reads nil.
Acceptance: `tests/lsp_dispatch_seams_acceptance.rs`, 14 tests, driven
against `pmacs_fake_lsp` through rust so nothing needs a toolchain.
Dispatch integrity is exercised at real co-occurrence — the fake server
writes `workspace/applyEdit` and the `executeCommand` reply back to
back, so both land in one `events_take` batch. 34b asserts affinity
survives a symlinked open and is paired with its own falsification: the
same resolver minus the canonicalize call spawns two servers, so the
positive case cannot be vacuous.
COHERENCE.md section 25 and the handoff/ledger update protocols make
these ride the PR.
COHERENCE.md:
- Section 6 named one optimistic key classifier and attributed it to the
GPU. There are two, one per replica frontend:
crate::optimistic::classify_key belongs to the pmacs --attach TUI
replica, and pmacs-gpu has its own unrelated optimistic_insert_text /
optimistic_crdt_insert. The section's "kept honest by
dispatch_idle_for" claim is confirmed for both, which this
investigation verified rather than assumed.
- Section 16 graded per-frontend degradation strong on the evidence of
per-frontend fold projection. That grade stands, but the practice is
enforced by convention rather than structure, and this defect is the
counter-example; the note says so and points at what is now structural.
- Section 2 step 8 records that the terminal was broken outright on the
GPU frontend, not merely undiscoverable.
docs/agent-handoff.md section 5 gains four lessons: adjacency does not
make two operations alternatives (and two individually sound idempotence
guards can be jointly useless); bite against every pre-image the fix
could have taken, since the obvious guard here fixes the storm and
introduces a controller leak; a quiet child is an instrument, because a
frame storm hides inside a chatty fixture and a geometric readout is
satisfied by an oscillating geometry; and TerminalMode::Raw makes
sh-based input fixtures useless because there is no ICRNL.
docs/active-work.md gains the lane entry with the branch, the bite
matrix, the named out-of-scope items, and the gate results.
Framing rev 7 adds S1-10..S1-12 -- the three findings that changed
behavior, each stated as the durable lesson rather than as a diff:
painting takes a buffer and seating takes the world, so any post-await
cursor operation needs an active-buffer guard; the rendered columns are
a contract Stage 3 is planned against, so precision yields to width; and
`open_directory`'s changed-nothing-on-failure invariant is itself a
probe, which is why the symlink descent no longer lists the target
twice. Plus the tolerant-channel note: cancellation was never a backstop
for a dired listing, because nothing cancels one.
The ledger records the round, the updated counts (dired 25 + 25 CRDT,
sweep 3,189 across 92), and the process lesson that cost me the fixes
once: a mutation-bite helper restores with `git checkout --`, so a fix
must be committed before it is bitten.
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.
Two review findings, both revision edits.
**Q#LN8's marker test was wrong in the other direction.** Rev 5 fixed
the directory case by reading a byte and requiring a non-nil read — but
an **empty** `lean-toolchain` reads nil at EOF too, so that rule
declines a marker that exists, silently, falling through to
`pmacs.project.detect`. Marker semantics here are `lean4-mode`'s
`locate-dominating-file` semantics: existence, not content, and a
`lean-toolchain` can legitimately be empty.
The discriminator is `read`'s second return, probed on LuaJIT 2.1:
| Path | `io.open` | `f:read(1)` | Verdict |
|---|---|---|---|
| file with content | handle | `"l"`, no error | marker |
| empty file | handle | `nil`, no error | marker |
| directory | handle | `nil`, `"Is a directory"` | decline |
| missing | `nil` | — | decline |
So `local data, err = f:read(1)`, declining only on a non-nil `err`. The
rule needs no per-platform re-probe: both directory behaviors are
declines, since a platform whose `fopen` refuses a directory fails at
`io.open` and one that opens it fails at `read`. There is no platform on
which a directory both opens and yields a byte.
Acceptance gains **24b** (an empty `lean-toolchain` marks a root) beside
24a, with the obligation that each be shown to fail against the
implementation satisfying only the other. A suite carrying just one is
satisfied by a resolver silently wrong for the other case — which is
precisely how rev 5's first answer got written.
**Citation sweep.** Round 4 stated the `project_root_for` correction in
§0.1 without editing the citation in §2.5; the correction and the fix
are different acts, and noting one is not doing the other. Review caught
a second stale citation (`handle_server_requests` at :1448), which
prompted a sweep of every `file:line` from §2.4 onward. Four more were
stale. All six: `project_root_for` 513 → 592, `ensure_server` 527 → 610,
`handle_server_requests` 1448 → 1549, `take_typed_edit` 12798 → 12827,
`pair.lua` 213 → 229, `compile.lua` 264 → 266. Six others were verified
good and left alone, listed in §0.1 so the next sweep knows what has
already been checked.
Q#LN15's present-tense "the change is small and spans two files" now
reads as past tense with its PR number, since that stage landed. Its
pre-#161 line numbers stay as written — historical record, not
navigation.
Reported symptom: text input in a terminal does not work on the GPU
frontend while the TUI is fine.
Root cause, measured against a real daemon, a real PTY child, and the
real pmacs-gpu attach client: the dispatcher loop applies BOTH
terminal-layout syncs to EVERY attached frontend each tick. A semantic
session satisfies both conditions -- it has a term_sizes entry from
AttachRequest and a semantic terminal declaration -- so the PTY is
resized twice per tick forever, because the grid arm runs the TUI
placement helper that subtracts a modeline the GPU never draws. The
result is a SIGWINCH storm plus screen reflow at tick cadence, which
makes an interactive line editor unusable while child output keeps
flowing. Measured 730 terminal frames in 20 s for a static screen.
Revision 2 answers Q#GT4 from the code rather than deferring it, which
changes the fix from a one-line guard into a split of
sync_terminal_layout: the grid arm is also the only per-tick
controller-liveness release a GPU session gets, and the semantic arm
neither performs it nor can be made to, because the buffer-follow
snapshot clears the viewport declaration that would drive it.
The doc also records three falsified hypotheses, why the Stage 3
real-path acceptance passes on the broken tree, and a COHERENCE.md
section 6 citation error (two replica frontends have two different
optimistic classifiers; the audit names only one).
No code changes in this commit.
docs/dired-framing.md rev 6: §0 gains the Stage 1 implementation notes
(S1-1..S1-9) -- the normalizer is exposed rather than mirrored (so B2 is
false by one small binding, in the direction Q#DR2 preferred); R2-3's
dedication claim is falsified by the display policy; acceptance 3c
cannot pin the descent routing and now says so; dired is the first
builtin to bind a mode-scoped key, which one pre-existing lib test
assumed impossible; `C-x d` takes no completion source on purpose;
ownership is the handle table alone; the mark column ships blank; a
symlinked directory needs a probe; and interactive origin does not
survive an await.
COHERENCE.md, per its §25 (an audited claim this PR changes updates
here, riding the PR): §1.1's interactive-file-opening fact, §2's journey
step 7, §4's beginner-level `files`, §14's tree bullet (Stage 1 landed a
flat listing and did NOT invent a tree convention), and §15's Priority 1
list. Step 3 stays **Missing at the CLI** with the mechanism spelled
out: `pmacs .` still exits 1, and this arc deliberately does not claim
the CLI path -- it supplies the buffer a directory should resolve to.
docs/active-work.md: the dired lane rewritten for Stage 1, including why
the branch is a fresh cut rather than a rebase of `dired`, the durable
substrate facts, the bite results (one VACUOUS, recorded rather than
relabelled), and the verification. Its canonical-base line was four
merges stale and now names 8c86d34.
docs/agent-handoff.md: one forward pointer only. The handoff describes
merged state, so it absorbs the substance when this merges.
Stages 1 and 2 landed (#160, #161). Re-scouting Stage 3 against `main`
@ `46a1b8f` — six merged PRs past the rev-4 snapshot — produced three
findings that change the plan and four that confirm it. Two were
established by running Lua in a fresh `EditorState` rather than by grep,
and are marked *probed* in §0.1.
**Stage 3 violated this document's own splitting rule.** §4 says "no PR
in this arc mixes a cross-cutting substrate change with Lean feature
content" and "a reviewer looking at Stage 3 sees only Lean" — while §4's
own risk column for Stage 3 read "two `lsp.lua` generalizations". Those
cannot both be true. One generalization shipped as Stage 2; the other is
Q#LN9's dispatch seams, which modify `handle_server_requests` —
confirmed the only production drain of LSP events, since
`LspManager::take_all_events` has no non-test caller. By the test that
justified splitting Stage 2 out, that is cross-cutting substrate. Stage
3 is now 3a (seams + canonicalizer, no Lean) and 3b (the Lean server),
strictly sequential.
**The Lean resolver could not satisfy the contract Stage 2 documented.**
#161 established that a configured root reaches `file_uri_for` verbatim
and that the resulting URI is the affinity key. Probed:
`pmacs.editor.file_path()` is not canonical — opening
`<tmp>/linkpkg/sub/./../sub/a.lean` through a symlink yields
`<tmp>/linkpkg/sub/a.lean`, lexical collapse only. No canonicalize
binding is exposed to Lua, and `pmacs.project.detect` canonicalizes but
returns nil without a marker. So one Lake package opened by two
spellings would spawn two `lake serve` processes — the bug Stage 2 was
built to prevent, re-entered through Stage 3's door. New Q#LN20 adds a
synchronous `pmacs.fs.canonicalize`; it rides 3a, and it serves every
future function-valued root rather than only Lean's. Two alternatives
are recorded with why they were rejected — the `detect`-anchored walk in
particular is incorrect, not merely inelegant.
**`pmacs.fs.stat` is unusable in the resolver.** It is async and the
resolver runs synchronously inside `ensure_server` ← `attach_buffer` ←
`buffer.after-load`, with no coroutine to await on. Probed: `io` and
`os` are exposed in the sandbox, so the marker walk uses `io.open` — the
opposite of what a reader would assume, hence Q#LN8 now says so. One
edge, also probed: `io.open` succeeds on a directory, so the walk reads
a byte rather than testing for a handle, and acceptance 24a bites the
version that does not.
Confirmed rather than changed: Q#LN7's stop-before-respawn is necessary
(default policy is OnCrash, the termination handler never consults the
exit code, and `maybe_restart` has no attempt ceiling — a broken `lake`
respawns forever; `stop()` setting `restart = Never` is what disarms
it); the response seam works as specified, since `Response` events are
pushed unconditionally and `send_request` returns the keying id.
One confirmation narrowed the design. `handle_server_requests` builds
its sid list from `attachments` and `push_event` is uncapped, so
subscribers fire only for servers with a live attachment. That turns
acceptance 34 into a reachable leak: killing the buffer with a request
outstanding strands the registration behind a drain that no longer runs.
The purge is now driven from both edges and 34 exercises the buffer-kill
path, which is the one a user can reach.
Also: §9 states the lane's coherence impact per COHERENCE §20 (journey
steps, interaction islands, config registry, background attribution),
including the honest note that 3b makes §2's step-3 grade marginally
worse by adding one more instance of the silent-spawn-failure class.
Three items are named in §6 rather than paid: the uncapped event queue,
the dropped `cfg.restart`, and surfacing the spawn failure itself.
Acceptance keeps every rev-4 number. The two split sections are
bulleted with literal labels because a markdown ordered list renumbers
from its first item, and 3b's criteria are non-contiguous; round 3's
finding 4 was stale references surviving a renumber, and not renumbering
is the cheaper way to not repeat it. Stale cross-references from the
split were reconciled in the same pass, and `project_root_for`'s
citation was corrected from 513 to 592 per COHERENCE §25.
`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.
Lean 4 Stage 2 (#161) landed while this branch's first-ever CI run was
in flight, which put the PR back to CONFLICTING at an unmoved head.
Merged rather than rebased, same as the 8c86d34 integration and for the
same reason: the PR is awaiting review rounds and a rebase would break
every review anchor.
The sole conflict was docs/active-work.md, as it was last time and for
the same structural reason -- every merge to main edits the lane ledger,
so a long-lived PR re-conflicts there and only there. Both sides' lanes
kept verbatim; main's updated Lean 4 heading taken over the stale one.
This integration is code-disjoint from the lane. Intersecting main's
changed files (COHERENCE.md, builtin/runtime/lsp.lua,
src/lua_bindings/mod.rs, tests/lsp_multi_root_acceptance.rs) against the
lane's own changed-file set leaves exactly docs/active-work.md, so none
of the first integration's pmacs-gpu/src/main.rs auto-merge risk recurs
here.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HZjWMjwPXhPbt9upku9mCk
Separates the pre-integration numbers, which described a tree 28 commits
behind, from the ones that describe what the PR now proposes.
The GPU test count is the integration proof rather than merely a pass: it
went 199 to 202, and e547a90 added exactly three tests to pmacs-gpu,
which is the entire delta on main since the merge base. Both sides' tests
are therefore present and running, and neither was dropped by the
auto-merge. Confirmed structurally as well -- main's fix survives as the
deferred closure form rather than the eager one that panicked, with its
regression test, alongside this lane's math work in the same file.
The lane was 28 commits behind. Merged rather than rebased, per the
#135/#137 precedent: the PR is awaiting review rounds and a rebase would
break every review anchor.
The only conflict was docs/active-work.md, where both sides add lanes.
Kept both: main's lanes verbatim, with this lane leading since it is the
one in flight. The conflict was pre-existing rather than introduced by
the dired or Lean 4 ledger commits -- it already conflicted against main
at e745068.
The integration surface, derived from git diff merge-base..main rather
than from another PR's file list, is pmacs-gpu/src/main.rs: main gained
72 lines there from e547a90, the minimap all-blank-slab divide-by-zero
fix, and this lane rewrites large parts of the same file. Git auto-merged
it textually. A clean auto-merge is not evidence the tree compiles, so
the full gate suite is what discharges it; the ledger records the
post-integration numbers separately from the pre-integration ones, which
described a tree 28 commits behind.
Rides this PR per COHERENCE.md §25 ("when a PR changes any audited claim
here, updating this file rides that PR").
§0 and §7: multi-root LSP affinity moves from in-flight-branch to PR
#161, and §7 gains the rule the slice actually establishes — a
*fallback* root is deliberately NOT an identity, so markerless files
keep sharing one server per language. That is the part a reader would
otherwise assume went the other way.
§1.2 gains the finding this PR turned up, which sharpens the audit
rather than restating it. The section recorded that background failures
produce "no `*errors*` entry"; the sharper fact is that **the channel
does not exist**. Fifteen call sites — `async.lua` (5), `syntax.lua`
(4), and one each in `lsp.lua`, `mcp.lua`, `fs.lua`, `editops.lua`,
`autosave.lua`, `commands/default.lua` — report through `pmacs.error`,
each guarded `if pmacs.error then ...`. It is defined nowhere in
production; the only assignment in the tree is a test stub
(`src/editor.rs:9881`), and `type(pmacs.error)` is nil in a fresh
`EditorState` (probed, not inferred). `pmacs.errors` plural in
compile.lua is an unrelated namespace. All fifteen are dead, and the
guard is what kept it unnoticed — it makes the silence read as
deliberate.
Hence the corollary now recorded beside the rule: report through a
channel with a **test that observes it**, or the guard is
indistinguishable from the silence it was meant to fix.
Also a frequency note: per-root affinity makes the
preconfigured-but-missing-server failure fire once per project root
rather than once per language per session. Unchanged in kind, strictly
more frequent. Surfacing it stays Priority 1 work needing its own
framing — it is user-visible product behavior (what message, where,
with what guidance), not a substrate fix to smuggle into an affinity PR.
Line-number citations in the touched sections re-verified per §25;
symbols are authoritative where they drifted.
COHERENCE.md §1.2 makes "a `pcall` around background wiring must log
attributed failure, never discard it" a standing rule, and names
`ensure_server`'s swallowed spawn failure as its canonical case — the
exact function this branch modifies. Round 1 deferred the resolver's
silent `pcall` as a Stage 3 concern. Under that rule it is not a
deferral, it is a fresh instance of the named anti-pattern added by a PR
touching the cited function, made worse by the memo: a raised error is
buried permanently for that directory and never observed again.
A resolver that raises, or returns a non-string non-nil, now leaves an
attributed trace naming the language and the directory. Returning nil
remains the documented decline and stays silent — pinned, so "report
failures" cannot be satisfied by reporting every resolution.
The report goes through `pmacs.editor.set_status`, NOT `pmacs.error`,
and that choice is the finding:
**`pmacs.error` does not exist.** Fifteen call sites across `async.lua`
(5), `syntax.lua` (4), `lsp.lua`, `mcp.lua`, `fs.lua`, `editops.lua`,
`autosave.lua`, and `commands/default.lua` report background failures
through it, each guarded `if pmacs.error then ...`. It is defined
nowhere in production; the only assignment in the tree is a test stub at
`src/editor.rs:9881`, and `type(pmacs.error)` is nil in a fresh
`EditorState` (probed, not inferred). `pmacs.errors` (plural) in
compile.lua is an unrelated namespace. So all fifteen reports are dead,
and the guard makes the silence look deliberate — which is why nobody
noticed. Writing the test is what caught it: the first version of this
fix used `pmacs.error` and its pin failed against a working
implementation.
Both bites recorded: dropping the report entirely fails the pin, and so
does reporting ONLY through `pmacs.error` — the dead-channel variant
this nearly shipped.
Not fixed here, deliberately: defining `pmacs.error`, the fifteen dead
sites, and surfacing the spawn failure itself. That last is Priority 1
work and a user-visible product behavior — what message, where, with
what guidance — so it needs its own framing rather than being smuggled
into an affinity PR.
Lands the approved dired framing on main as its own docs PR, and brings
the two required docs current after find-file merged as #162.
The framing was approved after two review rounds (seven findings, then
six) and revised twice more since: revision 4 recorded what implementing
Stage 0 falsified in the approved text, and revision 5 adds the coherence
impact statement that #163 made mandatory for every framing.
The coherence statement is new work, not a restatement. COHERENCE.md
section 20 Priority 1 already names this arc -- a find-file surface and
directory-argument handling -- so the framing now states which journey
steps it touches (7, and partially 3), that it adds no interaction island
because its keys are a mode-scoped keymap through the ordinary registry
and wdired is a mode swap rather than a modal layer, that it adopts the
config registry for dired.kill-when-opening, and that it inherits the
worker-attribution gap for its read_dir jobs without worsening it. It
also draws the boundary against the adjacent Journey Stage 1 arc: CLI
directory handling belongs there, the two meet at resolve_target_buffer,
and dired supplies the buffer a directory should resolve to rather than
growing a second directory surface.
One convergence worth recording: section 2 grades the golden journey
broken at step 3 because pmacs on a directory exits 1, and the mechanism
it cites -- File::open succeeding on a directory, then read_to_end
returning EISDIR -- is the same one Stage 0 pinned in its
accepting-a-directory test, where the pcall turns it into a status
message instead.
The handoff snapshot was stale through eight merges. It now anchors on
main at 2af1ab3, records COHERENCE.md as required reading and a required
framing input, and carries the two minibuffer facts find-file
established: a custom completion source cannot descend directories, and
a selected candidate shadows typed text -- both of which apply to M-x and
switch-buffer, not just find-file.
The ledger gains the dired lane with Stage 1's scope, the reason its one
Rust change cannot be done in Lua, and the rebase note for the dired
branch, whose framing commits become redundant when this lands.
COHERENCE.md states the product-coherence thesis (pmacs should be
immediately excellent, progressively understandable, completely
inspectable, and ultimately replaceable) and, per-section, the audited
ground truth of how the codebase measures against it: a scorecard
across 19 concerns, the golden-journey verdict table (breaks at "open a
real project" -- `pmacs .` exits 1), the six hardcoded key-interception
shadows with no transient-keymap mechanism to migrate them to, the
discoverability substrate-without-surface gap, the package/worker
identity gap, and three cross-cutting patterns (substrate without
surface, the silence asymmetry, per-arc coherence debt) that explain
most of the individual findings.
CLAUDE.md and AGENTS.md now list it as required reading alongside
agent-handoff.md and active-work.md, and ask new framing docs to state
their coherence impact. No runtime code changes.
Three of the five review findings land here; the other two are recorded
as named deferrals in the framing on the dired branch.
Finding 1: the command comment cited "acc4", a name from a draft scheme
that no test carries. It now names the real test, and the comment splits
the shadowing consequence into the two cases that actually exist -- a new
bare name that matches an entry (shadowed) versus one that matches
nothing (creates normally) -- each pointing at its test.
Finding 2: the everyday new-file flow had no test. Typing a bare name
that is not a subsequence of any entry is the path users hit first, and
the only route combining free text with a relative join; every existing
new-file test used a name containing a separator.
find_file_bare_new_name_creates_in_the_root covers it, asserting the
parent is the prompt's root so the join itself is pinned.
Finding 3: the failure arm was never exercised, and as the review noted,
deleting the pcall would have passed the whole suite. Accepting a
directory candidate reaches display_file, whose load fails because
File::open on a directory succeeds and the read returns EISDIR;
find_file_accepting_a_directory_reports_instead_of_raising pins that this
surfaces as the command's status message, leaves the active buffer alone,
and closes the prompt. Verified by manual revert: with the pcall replaced
by a direct call, that test and only that test fails. scripts/bite could
not isolate it, since the guard and its test have no separating commit.
Finding 4 is documented at the command rather than left implicit:
accepting on empty input opens the first-sorted candidate, because
fuzzy_score returns Some(0) for an empty needle and filter_and_sort
breaks the tie lexicographically, so dotfiles lead and a directory can
lead. M-x and switch-buffer share the mechanism, so it is inherited
rather than introduced, and it is listed in the framing beside the
accept-semantics change that would close it.
The blocker was process, not design. The test file was committed before
`cargo fmt` ran, so the reflow of five over-width assertions sat
uncommitted in the working tree while the branch as pushed failed the
first gate in CLAUDE.md. The "fmt clean" reported on the PR described
the worktree, not the branch. Gate results are only meaningful run
against the pushed tree, so this commit lands the formatting first and
the gates are re-run against it.
Two pins review asked for, each covering a branch the nine acceptance
tests left untested:
- A **string** `config.root` as an affinity key. acc17 covers only the
function form, so `return configured, "config"` had no test. The bite
puts both files in their own marked project: drop the config arm and
they key on their own detected roots and spawn two servers, so one
server on the configured root is only reachable if the override wins.
- `root = false` reads as unset. Defended by a truthiness check rather
than `~= nil`, previously by comment alone. Under `~= nil` the config
arm returns `false, "config"` and `file_uri_for(false)` returns nil, so
the file lands on a rootless server instead of its detected project.
Each was falsified against exactly the mutation it targets and neither
against the other.
Also documents an asymmetry review caught: `project_root_for`'s
"detected" arm is canonicalized for free because `pmacs.project.detect`
canonicalizes before walking, but a **configured** root — string or
resolver return — is fed to `file_uri_for` exactly as written, and the
affinity key is that URI. On macOS a resolver returning `/var/…` and a
detected `/private/var/…` are therefore different keys for one
directory, silently yielding two servers for one project. There is no
Lua-side canonicalizer to normalize it, and Stage 3's Lean resolver is
the first real consumer, so the obligation is stated in the
`config.root` doc comment where that resolver's author will read it.
Stage 1 merged as #160 (`main` @ `0827dd1`); the Lean lane header and
branch line now say so, and Stage 2 gets its own subsection.
Edits stay inside the Lean lane. PR #156 is still open against both this
file and `docs/agent-handoff.md`, and it rewrites the snapshot header,
the canonical-base line, and the whole bottom-panel lane — so those are
left alone rather than merged twice. `agent-handoff.md` is untouched for
the same reason plus its own: §1 describes what is on `main`, so it
updates at merge, not during review.
Records the one finding this stage turned up but did not fix:
`ensure_server` never forwards `cfg.restart` to `pmacs.lsp.spawn`, so a
`restart` in `pmacs.lsp.config[lang]` is silently dropped on the
auto-attach path. Pre-existing, and out of scope for a PR whose
acceptance 16 pins existing attach behavior as unchanged.
Dired arc Stage 0 (docs/dired-framing.md section 10, Q#DR11). Until now
pmacs had no discoverable way to open a file by path: no find-file
command and no C-x C-f binding, so a file entered a session only from
the CLI, an LSP jump, a project-search visit, or C-x C-r, whose prompt
does pass free text through but completes only over the recent list.
The command prompts with completion rooted at the active buffer's
directory, or the process cwd when the buffer has no backing path, and
opens the result through pmacs.window.display_file. A path that does not
exist yet creates a buffer bound to it with the "[new file]" status,
which is Emacs parity and comes from resolve_target_buffer rather than
anything added here. Nothing is written to disk until the user saves.
Two substrate facts shape the design and are documented at the command
rather than left to be rediscovered.
Completion is flat: the files source lists one directory and yields bare
basenames, and a custom function source could not do better, because
sources are called with no arguments and run synchronously outside any
coroutine, so a callback can neither see the input to re-root on nor
await a directory listing. Hierarchical completion is a named Rust
change in the framing.
A selected candidate shadows typed text: recompute_candidates selects
index 0 whenever the candidate list is non-empty, and
resolve_accepted_value returns the candidate over the typed contents. So
typed text reaches the accept handler exactly when the input filters
every candidate away, which for basename candidates under a subsequence
filter means when it contains a separator. That makes the deeper-path
case work verbatim and leaves one hole: a new bare name that is a
subsequence of an existing entry opens the existing file. The acceptance
pins that as a decision rather than an accident; closing it needs a Rust
change to accept semantics that Stage 0 deliberately does not make.
A leading tilde is expanded before the path reaches the core, because
get_or_load_buffer normalizes the path it stores but loads from the raw
one -- so an unexpanded tilde path deduplicates against an already-open
buffer yet fails to load a file that is not open yet.
The prompt field starts empty and names its root in the prompt string
instead: any prefill would contain a separator and silently disable
completion.
Acceptance is dispatch-driven throughout -- a real C-x C-f, real typing,
a real RET -- so a dead binding cannot pass vacuously and the Lua
lifecycle accept(), which bypasses the path interactive input takes, is
not used.
`ensure_server` reused any live server whose `language_id` matched,
regardless of project root — its own comment documented this as a known
post-v0.1 limitation. For project-model-strict servers that is a
correctness failure, not a rough edge: `lake serve` is bound to one Lake
package, rust-analyzer and gopls to one workspace, so the second project
a user opens gets a server that cannot resolve its imports.
Server affinity is now keyed on the project root, with one rule that
keeps the change from regressing every other language:
The affinity key is the root only when a root was actually FOUND.
`project_root_for` never returns nil for a file that has a path — its
last resort is the file's own directory — so a naive `(language_id,
root)` key would give every directory of loose scratch files its own
server, for every language: two stray .py files in different directories
would spawn two pyrights where today they share one. It now returns
`root, source` with source one of "config" / "detected" / "fallback",
and only the first two become an affinity key.
Matching is on the spawned spec's `root_uri`, nil matching nil, so the
fallback spawn passes `root_uri = nil` for the key and the stored spec to
agree. `cwd` still carries the directory, and `build_initialize`
(src/lsp.rs) derives the identical `rootUri` from `cwd` when the field is
None — using a percent-encoder with the same allowed set as Lua's
`file_uri_for`. The initialize payload for that case is therefore
byte-identical to before; only what the reuse loop matches on changes.
`build_initialize` is the only reader of `spec.root_uri` in the tree.
Two consequences, both deliberate and both asserted rather than
discovered:
- A server hand-spawned from init.lua with only `cwd` set also reads
back nil, so a root-bearing attach will not adopt it. We cannot know
which root it was meant to serve, and guessing wrongly routes a
project's files to the wrong server.
- Opening files across N project roots spawns N servers. rust-analyzer
has the same property and no editor caps it by default; `pmacs.lsp.stop`
is the manual escape and an LRU reaping policy stays deferred.
`config[language].root` may now be a `function(path) -> string|nil` as
well as a string, for languages whose root rule the shared marker walk
cannot express — an innermost-wins walk cannot find an *outermost*
marker. A resolver returning nil declines and falls through to the marker
walk. Results are memoized per directory because hoisting the root
computation above the reuse loop puts it on every attach rather than
every spawn; the memo is keyed weakly by the resolver function itself, so
replacing `config[lang].root` cannot serve a root the old one computed.
`pmacs.lsp.list()` rows gain `root_uri` and `cwd`. `root_uri` is the spec
field verbatim, deliberately not the URI the server was initialized with.
No protocol change. No Lean content: this is the shared affinity function
for every LSP language, so it ships as its own PR and is exercised
through rust, python, go and typescript against `pmacs_fake_lsp`.
tests/lsp_multi_root_acceptance.rs covers acceptance 13-21. Every fixture
sets `pmacs.project.set_search_boundary` at its own tempdir root:
without it the marker walk climbs to the filesystem root, and a stray
`.git` above the temp directory would turn the markerless cases into
detected ones — the assertions would still pass while testing nothing.
Refs docs/lean4-mode-framing.md Q#LN15, acceptance 13-21.
Review round 1 flagged that neither ledger knew about this branch, and
`docs/active-work.md`'s stated job is exactly the volatile open lanes.
Records the branch, base, framing revision, what Stage 1 ships, the
discharged Q#LN1 obligation, the Q#LN4 blast radius, and the four
implementation findings that are not in the framing (the `warning`
colour collision with `number`, `Some(1)` resolving to `@function`
rather than `@constructor`, the `module > declaration > def` nesting,
and `injection_aliases` being a write-only proxy). Also carries forward
the two Stage 2 corrections the framing already holds, since that lane
starts next.
Deliberately ADDITIVE ONLY -- one new section, zero deleted lines. PR
#156 is open against both this file and `docs/agent-handoff.md` and owns
the snapshot header, the canonical-base line, and the bottom-panel
lane's status. Touching those here would collide with a PR already in
review, which is the "frozen reviewed PRs do not absorb moving
overlapping work" lesson from #135/#137.
`docs/agent-handoff.md` is deliberately untouched: its §1 snapshot
describes what is ON `main`, so it gets updated when this merges, not
while it is in review.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Review finding: acc12's server-list assertion could not fail for the
regression class it was written to catch. The shared `editor()` helper
runs `pmacs.lsp.config = {}` before any buffer opens, so
`#pmacs.lsp.list() == 0` holds for every language regardless of what
Stage 1 ships -- a Stage-3 front-run that added
`pmacs.lsp.config.lean4` in a builtin runtime file would have slipped
straight past it. The same vacuous-assertion shape as #155 R2.
acc12 now asserts the actual claim against a PRISTINE `EditorState`,
before any config wipe: no builtin runtime file defines
`pmacs.lsp.config.lean4`. A non-vacuity check pins that the same lookup
finds `pmacs.lsp.config.rust`, so this cannot pass merely because the
table is empty or absent.
Bite-verified: adding `pmacs.lsp.config.lean4 = ... { command = "lake",
args = { "serve" } }` to `builtin/runtime/lsp.lua` fails the test; the
stub was reverted.
The process-list half is kept and its comment now says why it survives
the wipe: a direct probe spawn from a future `lean.lua` shows up there
whatever `pmacs.lsp.config` contains.
Also fixes a stale column in a `highlight.rs` comment -- the Lua table
brace in `local t = {}` is at col 10, which is what the code already
used.
Gates rerun: fmt and strict workspace clippy clean; 1,826 default +
2,003 CRDT library tests; lean4 Stage 1 9/9; M4 121; required GPU 152;
isolated-config workspace sweep 3,150 across 90 suites; diff check clean.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Levi Neuwirth