pmacs/docs/active-work.md

Active work — cross-machine resume ledger

Snapshot: 2026-07-29. This file records volatile work that has not landed on main. Read it after docs/agent-handoff.md. Remove completed entries when their PR merges; do not let this become a second permanent backlog.

No lane below is retained past its merge. This snapshot removes the resource-op delete guard (#186) and dired Stage 2 framing (#171) lanes the moment their PRs merged, because the same commit put their load-bearing decisions into docs/agent-handoff.md §1 — rule 4's precondition, satisfied deliberately rather than deferred. The bottom-panel lane is not removed: 2B-2 landing leaves 2B-3 and Stage 3 ahead of it, so the lane is rewritten to the remaining plan.

#188's lane arrived with #188, which is the point: with several PRs open, a lane written on main for work that lands elsewhere re-conflicts on every merge. Written on its own branch it costs one conflict, at the merge that would have happened anyway.

The PTY terminate diagnostic (#176) was the last lane retained past its merge — retained because rule 4 removes a merged lane only after its durable facts reach docs/agent-handoff.md, and that absorption was unowned. The 2026-07-28 snapshot owned it: #176's facts are now in the handoff (§1's arc bullet and §5's two ops lessons about ticking observers and proving child exit), so its lane is gone. The Lean 4, GPU-terminal-input, inline-math (#172), dired (#169), and terminal config + copy mode lanes were removed the same way — the last of these was #180's work, folded into #182 so two open PRs would stop re-conflicting in this file.

Trust the canonical-base line below over any lane header: if a PR number appears in git log --first-parent githubsucks/main, it has landed regardless of what a lane says.

Two open PRs had no lane here at all before the 2026-07-28 snapshot — #174 and #171. An open PR is exactly the volatile work this file exists to record, so its absence is a ledger defect rather than a tidy omission: #171 drifted 153 commits while invisible here, and its still-green old CI run described a tree nobody had looked at since. When a PR is opened, give it a lane. All three have since merged — #174, #171 and #186 — so per rule 4 their lanes are gone again and their durable facts are in docs/agent-handoff.md (§5 for #174's lesson, §1 for the two framings).

Repository authority

• Canonical development URL: https://github.com/levineuwirth/pmacs.git. This ledger uses the normalized local alias githubsucks so its refs and recovery commands are identical on every machine. Remote names are otherwise machine-local: origin may name this canonical URL, a release mirror, or something else, and therefore has no authority by name alone.
• Canonical base at this snapshot: githubsucks/main @ 7586905 (the docs-only coherence listview correction #189, atop the docs-only landed-state refresh #185, the M4 config-sink race fix #174, bottom-panel Stage 2B-1 #184, the Journey/GPU directory-target ratchet #183, Journey Stage 1a #182 and the previously recorded landed work). Protocol schema support is v6..=v21; the production server-first Hello still advertises v20 — two different facts, and #184 landed only the first. The previous snapshot named 0442d78, and the recovery floor advances with it: the check below now requires 7586905 or newer, so a tree at 0442d78 no longer passes. That is deliberate — the floor moves with the base, because a check that accepts an older commit than the declared base passes on a tree the rest of this file does not describe. Lanes below that name an older base have not been re-based; derive their integration surface from git diff <their base>..main.
• On the transfer source, origin/main named a release mirror at d3fa632 and lagged badly. On the current destination, origin names the canonical URL. This difference is why all recovery begins by verifying URLs and normalizing githubsucks rather than trusting origin/main.
• The shared desktop checkout contained unrelated uncommitted work. The branches below were prepared in isolated worktrees; never clean or overwrite the shared checkout to recover them.

Start on another machine by inspecting its remotes:

git remote -v
git remote get-url githubsucks

If the second command says the alias is absent, add it; if it prints a different URL, stop and resolve that collision rather than overwriting an unknown remote:

git remote add githubsucks https://github.com/levineuwirth/pmacs.git

Then recover current refs:

git fetch githubsucks --prune
git log -1 --oneline githubsucks/main
git worktree list
git status --short --branch

The git log command must expose 7586905 — the base named above — or a newer intentional main. Keep this threshold and the canonical-base line in step: a recovery check that accepts an older commit than the base it declares canonical will pass on a tree the rest of this file does not describe. If it does not, stop and repair the remote/fetch configuration.

The CRDT half of the test corpus is dark in CI — NEEDS A LANE

• No branch, no framing yet. Found while gating #166, then measured properly during the vterm as-framed audit. Deliberately kept out of #166 so a CI change would not arrive after review approval.

• Root cause: .github/workflows/ci.yml never enables the crdt feature anywhere — zero hits across the workflow directory. The test job runs cargo test --all-targets --no-default-features --features luajit|lua54. Every #[cfg(feature = "crdt")] test is therefore not compiled in CI, not merely skipped.

• Measured, --list under CI's exact flags versus the same flags plus crdt: 3,176 vs 3,449 — 273 tests dark. Re-measured at 74301d1 (2026-07-26; at fe8b8ba it read 3,170 vs 3,443, the same 273 dark — #176 added six tests, none of them crdt-gated). The number moves with every merge and must be re-measured, not quoted. #168 reported 3,024 vs 3,288 — 264 dark, 177 in the library — at 1b6a084; #178 then added CRDT-only generated-buffer coverage, and other lanes landed CRDT tests in between. Per target:

  dark	CI	full	target
  185	1,848	2,033	the library itself (src/lib.rs)
  21	15	36	m5_5_acceptance
  13	1	14	gpu_invocation_acceptance
  13	1	14	gpu_initial_target_acceptance
  8	0	8	m10_11_acceptance
  6	0	6	auto_pair_crdt_acceptance
  6	0	6	m10_2_perf
  4	5	9	vterm_stage3_acceptance
  4	0	4	m10_10_perf
  3	0	3	compile_mode_crdt_acceptance
  2	22	24	theme_faces_acceptance
  2	0	2	m11_5_semantic_acceptance
  1	14	15	terminal_copy_mode_acceptance
  1	9	10	vterm_stage1_acceptance
  1	7	8	statusline_segments_acceptance
  1	10	11	gpu_font_acceptance
  1	0	1	auto_indent_crdt_acceptance
  1	0	1	m10_11_perf

  The rows sum to 273; the table is the whole census, not its head.

• The single worst line is the library. cargo test --lib --features crdt is a REQUIRED local gate in CLAUDE.md, and CI has never run it. 185 library tests — the whole CRDT half — are developer-machine-only, and that count grows with every merged branch that adds a crdt-gated unit test.

• Ten suites run zero or one test in CI, including gpu_initial_target (#148's entire acceptance, 1/14), gpu_invocation (#141's, 1/14), and a37, the Vterm Stage 3 real-daemon/real-PTY/real-wgpu path that #135 built specifically because "a decoded-message fixture would prove none of the three fit together".

• ⚠ a37 will report green in the new job without running, unless the job builds pmacs-gpu AND sets PMACS_REQUIRE_GPU=1. Measured 2026-07-26 while gating #173. a37_real_daemon_real_pty_and_headless_gpu_ render_one_terminal_session derives its sibling binary path from CARGO_BIN_EXE_pmacs, and on a missing binary it eprintln!s a skip and returns ok. A fresh worktree running cargo test --features crdt --test vterm_stage3_acceptance reports 9/9 in 0.17 s having never run it; a real run takes ~4 s. Only PMACS_REQUIRE_GPU=1 promotes that skip to a failure, and CLAUDE.md applies that flag to cargo test -p pmacs-gpu — a different package, so the required local gate does not cover a37 either. The gpu-render job already sets the flag, which is what makes fix-shape part 2 sound; state it as a requirement of that job rather than inheriting it by luck, because a crdt leg added to the plain test job would run a37 vacuously.

• a37 is also load-sensitive, which changes how to read the expected first-run failures. It passed at d152120 and failed at that same commit twenty minutes later, with a second agent saturating the machine with rustc in between; it then failed identically on d152120, 04c5ad1, and the #173 merge commit, which is how #173 established the failure was not its own. The signature is last_frame_text all spaces with rendered_nonuniform_frames nonzero — frames arrive, content does not. pmacs-gpu's own suite flaked the same way under the same load (201/202, then 202/202 on immediate rerun). So a red a37 on the first CI run is ambiguous by construction: before treating it as a real failure, run the same command on the merge base, and prefer serialized execution for this suite over retry-until-green.

• Sort deliberate from accidental before proposing a fix. Some of the 264 are perf suites that are #[ignore]d by default and belong to their own jobs (m10_2_perf 6, m10_11_perf 1). m10_10_perf has no #[ignore] and no CI job naming it, so it looks accidental. This classification is not finished and is the lane's first task.

• Fix shape, two parts (the flag combination is verified to work: --no-default-features --features luajit,crdt lists 10 vterm Stage 1 tests versus 9 without):

  1. a crdt leg on the test job for the non-GPU suites and the library;
  2. the GPU-requiring crdt suites onto the existing gpu-render job, which already has lavapipe and PMACS_REQUIRE_GPU=1 — vterm_stage3_acceptance, gpu_invocation_acceptance, gpu_initial_target_acceptance, gpu_font_acceptance.

• Expect first-run failures, and budget for them. These would execute in CI for the first time ever: real PTY timing on CI runners, wgpu under lavapipe, and daemon-socket tests at unfamiliar concurrency. Start ubuntu-only and decide about macOS from evidence. A red first run is the lane working, not the lane failing.

• Mitigating fact, verified rather than assumed: #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.

• This lane also owns a --lib --features crdt flake, observed and scoped without overclaiming its cause (inherited from #178's gating, where the terminal lane recorded it). cargo test --lib --features crdt failed ~1 run in 5 on process::tests::setsid_escapee_is_not_reaped_and_teardown_reclaims_readers — active_reader_probe returning None at process.rs:3179 ("live runtime probe"). Pre-existing and unrelated to #178: that branch did not touch src/process.rs at all, and the test passed 10/10 standalone; the observed failures were during parallel full-suite runs. That localizes the trigger to suite load or interaction, but does not distinguish parallelism from another full-suite effect — no serial full-suite bite was run. The leading code-path explanation is the known drain_until trap: draining for Started also ticks, and a tick can reap the leader before the following active_reader_probe. That is an inference from the failure site and control flow, not yet a falsified root cause. Discriminating it belongs here. Two unnamed CRDT failures in #178's round-2 gating are a plausible match but remain unattributed — no test names were captured.

• A second standing obstacle for this lane: cargo clippy --workspace --all-targets --features crdt -- -D warnings fails on main — measured at 74301d1: seven errors before the build aborts, four in src/daemon.rs (useless_conversion at 3996, missing doc backticks at 4076, too_many_lines 112/100 at 4083, an unneeded mut at 4965) and three in tests/vterm_stage3_acceptance.rs (too_many_lines at 637 and 793, a redundant continue at 843). Treat that as a lower bound, not an inventory: Clippy abandons the remaining targets once one fails, and a run on an older tree surfaced a further doc-backticks error in tests/auto_indent_crdt_acceptance.rs:42 that this run never reached. The standing gate list runs Clippy without crdt, so these lints have never been enforced. Any CI job that compiles the crdt targets has to fix them first or it will be red on arrival.

Generated-buffer immutability lane (Arc: workbench primitives) — STAGE 1 OPEN

Framing: PR #188, revision 7, approved and merged to main as 27b1185. #188 owns the acceptance contract; this lane adopts it. On 2026-07-29 the user directed #191 to fold its review corrections into this branch and then merged #188, settling the implementation authority and merge ordering. The contract is now docs/generated-buffer-immutability-framing.md on canonical main.

• Branch generated-buffer-immutability-stage1, worktree ../pmacs-gbi-stage1. githubsucks/main is integrated into it. Measured when this line was written:

  $ git rev-parse --short githubsucks/main
  27b1185
  $ git log --oneline -1 githubsucks/main
  27b1185 Merge pull request #188 from levineuwirth/generated-buffer-immutability
  $ git merge-base --is-ancestor githubsucks/main HEAD && echo "main IS integrated"
  main IS integrated
  

  That is a reading, not a constant, and it went stale inside this lane's own review round. main moved four times while the lane was open: #187 -> #192 -> #193 -> #188. An earlier revision of this bullet pasted the same three commands with 64883eb and the same main IS integrated line, and #193 merged between writing it and pushing it --- so the pasted output was false in the tree that carried it. Pasting command output is necessary and not sufficient: re-measure at push time, and treat any base SHA in this file as expired on sight.

• What Stage 1 ships. dired.lua's paint and listview.lua's render write through pmacs.buffer.set_generated_contents (zero bypass_intercept writes remain in either file); listview gains Q#GB13 ownership-by-handle with <2>..<99> disambiguation and Q#GB18's identity-routed panels list in the same commit; Q#GB6's cursor/view-top clamp plus selection clamp-or-clear in both EditorCore::notify_buffer_edit and rebuild_views_for; listview refresh reseating through the already-notified view rather than a redundant same-buffer switch; and Q#GB16(a)'s corrected fold status string. No protocol change, no new Lua surface, no new interaction island.

• Why these two families first, and it is not "the cheap half". compile.lua:219 and builtin/commands/default.lua:855 rebind all seven undo chords to a no-op; dired.lua and listview.lua rebind nothing, so a bare C-/ emptied a listing and a panel. Stage 1 closes the only two families reachable without M-x.

• Review round 1 found the stale selection anchor and four acceptance contract mismatches. Its provisional drop-on-stale fix stopped the crash but intentionally waited on #188 to decide the selection rule; criteria 5 and 7 likewise recorded evidence without claiming to replace the framing. That evidence produced #188 revision 7.

• Review round 2 closes both remaining P1 findings against revision 7.

  • Q#GB6 now matches at both sites. Cursor and anchor clamp to the new extent; a selection survives shortened unless an endpoint movement collapses it, in which case it clears. acc16h and acc16i each drive a real caller and assert both the surviving region and collapsed case. Unconditional drop and bare clamp are separately falsified.
  • The Stage 1 criteria are adopted without local substitutes. Criterion 5 has the exact rope-refusal + byte-identity half and the Rust-lifted named-intercept half for both adopters. Criterion 7 now bites the named fan-out mutation for both adopters: listview refresh no longer rebuilds the view with a redundant same-buffer switch. Criteria 11 and 12 carry the framing's [main] classification and also record where its narrower Q#GB13-without-Q#GB18 pre-image fails.

• Stage 2 still owes everything with new Rust in it, per the framing's cut: Buffer::apply_generated_edit + GeneratedOutcome + the { generated = true } option + its own run_buffer_edit arm; set_generated_contents reimplemented over it; Q#GB10's path-backed refusal and mark_clean; Q#GB15's identity_protected; Q#GB13/GB18 for compile.lua and the search panel; Q#GB5's ensure_slot lock; conversion of the remaining 13 write sites; and the three compile_mode_acceptance intruder tests converted per Q#GB12.

• Verification at code checkpoint 5d92348. The ledger commit on top is docs-only; cargo fmt --check and git diff --check are re-run after it. cargo fmt --check clean; cargo clippy --workspace --all-targets -- -D warnings clean; library 1,863 passed + 3 ignored default and 2,048 passed + 4 ignored CRDT; listview_acceptance 17, dired_acceptance 31, folding_acceptance 21, terminal_copy_mode_acceptance 18 default and 19 with --features crdt — judge that step by the count, because acc16e is #[cfg(feature = "crdt")] and a default run never compiles it; M4 121 passed + 3 ignored + 1 filtered with --skip basedpyright; required GPU 202/202. The first GPU attempt inside the tool sandbox failed three managed-attach socket tests and left the closed-outbox reader blocked; the authoritative rerun outside that socket sandbox passed all 202. git diff --check clean.

• The dired 200 ms perf test is load-sensitive, and the conversion costs it nothing. Review saw dired_renders_10k_entries_within_200ms take 241 ms in a combined run and pass alone. Measured here: 0.09 s isolated over five runs, and the whole 31-test suite finishes in 0.12 s, so 241 ms was contention rather than a regression. Measured against the pre-image as well, by swapping in main's dired.lua (the bypass_intercept paint): 0.09 s either way over five runs each. A whole-buffer set_generated_contents costs the same as the bypass replace it replaces, which discharges Q#GB4's measurement obligation for the whole-buffer case only — the streaming case is Stage 2's and is not touched here.

• Bites, re-run under scripts/bite's positive control (#192). A bare bite: OK from the pre-#192 script is weaker than it looks, so every result below is from the current script or from a mutation harness carrying the same control (named tests must pass on the working tree and at least one must have run).

  • Falsified by revert, all OK (assertion) — not OK (COMPILE): builtin/runtime/listview.lua for criteria 1, 2, 9 and 10; builtin/runtime/dired.lua for criteria 3 and 13a; src/lua_bindings/fold.rs for 13b; src/editor_core.rs for 8, 8b and both selection-normalization pins.
  • Falsified by a named mutation, each observed to fail: the fan-out drop in the set_generated_contents binding (criterion 7); deleting self.read_only = false (criterion 4, both adopters); deleting add_intercept and set_round_trip_input at each adopter (criteria 5 and 6); the name-keyed panel_for_buffer (criteria 11 and 12); adopting at the variant limit (criterion 10); the old fold status string (13b); deleting each clamp (8, 8b); deleting the selection helper from either site; unconditionally dropping a stale anchor; and retaining a selection that an endpoint clamp collapsed. Criterion 7's fan-out drop now fails by assertion in both listview and dired.

• Recovery:

  git fetch githubsucks
  git worktree add ../pmacs-gbi-stage1 generated-buffer-immutability-stage1
  cd ../pmacs-gbi-stage1
  cargo test --test listview_acceptance --test dired_acceptance
  cargo test --test terminal_copy_mode_acceptance --features crdt
  

Bottom-panel lane (Arc 7) — 2B-3 OPEN; Stage 2 is COMPLETE with it

Stage 1, the Stage 2 framing, Stage 2A, Stage 2B-1, and Stage 2B-2 are all on main. Stage 2B-3 — the GPU panel band, compatible protocol-v21 activation, and the negotiated panel_capable flip — is this lane's open work, and it completes Stage 2. Framing revision 5's three-way split of 2B was explicitly approved on 2026-07-27; revision 6 records PR #184's review correction.

• Branch bottom-panel-stage2b3, worktree ../pmacs-bp-2b3, cut fresh from landed main @ e003b81 (Merge pull request #190 from levineuwirth/resource-op-delete-guard-impl) — measured with git log --oneline -1 githubsucks/main, not carried over from a briefing. Not stacked on ../pmacs-bp-stage2b2: each slice starts from landed main, which is the house pattern and what 2B-2 did.

• The activation mechanism, and why the alternatives fail. The constraint inherited from 2B-1's review is that the unsolicited Hello may not become 21: it is server-first, so a shipped v20 frontend rejects an unknown version before it can send AttachRequest. What 2B-3 ships instead is a frontend counter-offer: ADVERTISED_PROTOCOL_VERSION becomes a permanent compatibility baseline (20), the frontend answers requested_protocol_version(baseline) — its own PROTOCOL_VERSION when the baseline is current, a verbatim echo of anything older — and the daemon records negotiated_session_version(offer). The daemon needed no change to accept it; it already recorded req.protocol_version. Rejected alternatives, each for a reason the server-first shape forces: growing Hello/InstanceCapabilities (postcard structs are positional, so it breaks the other direction symmetrically and cannot be verified against shipped binaries); a client-first hint (an old client writes nothing until it has read Hello, so any probe is a timeout); sniffing instance_identity.pmacs_version (version-string branching on a field documented as display-only); and a second post-Hello daemon message (an old client reads it as an InstanceMessage and dies on the unknown discriminant).

• The window this leaves open is named, not hidden. A daemon whose own PROTOCOL_VERSION equals the baseline rejects an offer above its supported range. A single AttachRequest cannot mean both "I want 21" and "≤ 20", so compatibility is preserved for old frontends — the direction that matters, since the daemon is what a user leaves running — and the window closes on the next daemon restart. It surfaces as an explicit GoodbyeReason::VersionMismatch naming both versions, pinned by an_unsupported_offer_is_refused_by_name.

• Review round 1 closed six findings at PR #198, four of them one shape. The panel layer was a partial port of the document/terminal layer and the new tests asserted the declaration side only, so each omission was invisible. Audited as a port rather than patched as a list, which found two more gaps of the same shape that review had not named: the headless probe never armed the panel wire at all (so nothing could ever exercise a band), and a disconnect left the band on screen.

  • Geometry agreed in one place instead of three. The declaration used the stable probe while painting and hit-testing used the document-dependent mono_advance, and the declaration inset its width by TEXT_LEFT against the parent framing's explicit x=0 full-width contract. The advance is now cached behind the declaration (PanelBand::declared_advance), so the three cannot drift: there is one value. Framing rev 7 R7-1/R7-2 records the widened contracts.
  • One classifier owns "does the band claim this pixel". Four pointer handlers each decided for themselves and three simply did not ask, so right-click and wheel were applied to the document underneath and a held left button was reported as a hover. PointerSurface / classify_pointer_surface makes forgetting the band impossible to do quietly, and makes the routing testable without a window or a daemon.
  • focused decides the caret. The producer ships cursor for a passive panel too, so painting it unconditionally put a second insertion caret on screen.
  • Planned underlines were never consumed. Straight forms now ride the quad batch, curly the squiggle pipeline — the terminal path's split.
  • VersionMismatch { server } is the instance's PROTOCOL_VERSION. Reporting the advertised baseline told a frontend the daemon topped out at 20 while it spoke 21. The acceptance had pinned the wrong value, so the test was holding the bug in place; fixed and re-pinned.
  • Criterion 54 is satisfied, not caveated. The panel-hosted probe is a54_real_daemon_real_pty_and_headless_gpu_render_one_panel_hosted_terminal — one real daemon, one real PTY child, real wgpu, through display = "panel". Opening a terminal and moving it was the first attempt and is subtly wrong: the buffer ends up displayed twice and the document window keeps projecting it, so the acceptance cannot tell a panel-hosted child from a document one.

• Round 1 verification: 41 bites, 41 OK. The original 23 re-run after the round-1 changes (none went vacuous) plus 18 new ones covering every changed pin. The forbidden act is bited twice: mutating the constant trips a const-block assertion at build time (stronger outcome, weaker evidence), so a narrowed swap that compiles both ways mutates the daemon's Hello value directly. A probe bite must rebuild pmacs-gpu — cargo test does not rebuild the sibling binary the acceptance runs as a process, and without that the mutation never reaches it and the bite reads as vacuous for the wrong reason.

• What 12/12 green did not catch. The reviewer confirmed all twelve checks, bottom_panel_stage2b_gpu_acceptance 4/4, and required GPU 216/216 on the head that carried all six defects. The gate suite is not the check on a slice like this; the assertions are.

• Six commits: the activation mechanism, the three-boundary split plus the band's machinery, the paint/input wiring, the acceptance suites, the bite fixes, and the round-1 port completion — plus docs.

• Verification on the reviewed head (all redirected to files and checked by their own exit codes, never through a pipe): recorded in the PR body and reproduced by the commands in CLAUDE.md.

• 23 falsifying mutations, all executed, all biting, using an in-place mutation runner rather than scripts/bite because several claims live in the same file as their test and a whole-file swap would not compile. Three assertions were VACUOUS on the first pass and the mutation runs are what found them — a fixed-point contrast check, a pixel test that passed with the band painting nothing, and a monospace fixture that could not tell two glyph-advance derivations apart. A fourth finding was about the code, not the tests: "no panel frame reaches a v20 session" is defence in depth, not the placement gate, and passes with the capability gate removed entirely.

• Two real defects the new tests caught in the implementation, both recorded in docs/agent-handoff.md §1: edge_scroll_direction has no upper bound, so reclassifying its boundary was necessary but not sufficient; and apply_panel_payload ignored the exhaustion latch.

• Cross-machine recovery (fresh clone):

  git fetch githubsucks --prune
  git switch --track -c bottom-panel-stage2b3 githubsucks/bottom-panel-stage2b3
  git rev-parse HEAD
  

• Stage 3 (the adopter default flip) is the arc's last step and is not started. This lane is not removed at 2B-3's merge: Stage 3 remains ahead of it.

• Stage 2B-2 MERGED as #187. One review round of five findings on top of the implementation, 12/12 green, 22/22 mutations biting. Its durable lessons are in docs/agent-handoff.md §1.

• What PR #187 shipped, dark by construction: the semantic daemon's FrontendCellGeometry epoch machine; one reconciled panel grid derivation; PanelFrame::{Present, Absent} projection on both document and terminal semantic paths; stable presentation epochs; resize and pointer validation against the live window/buffer/epochs; the panel's own statusline context; and pre-drain semantic panel-terminal resize. It does not add the GPU consumer or enable the capability.

• Review round 1 closed five findings plus one sweep result at 3ecb03d. The wire-area clamp became durable hide state; a stale same-buffer reopen can no longer retain input authority; semantic panel terminals resize before child drain; NoMessage retains a published band baseline while Invalidated clears it; wheel activation follows the terminal-only focus rule; and legally wide panels clamp their PTY content without disappearing.

• Review round 2 closed two findings at bfaaf2b plus this ledger commit. Side affinity can replace the buffer while preserving the WindowId, so retained panel statusline segments are now keyed by the full (WindowId, BufferId) presentation. Every authoritative Absent also clears that baseline, including duplicate-suppressed Absent, so a later Present under NoMessage cannot resurrect peer state that was already cleared. Two acceptance tests bite those exact transitions. This lane and docs/agent-handoff.md now name the open PR, current landed base, checkpoint, and 2B-3 ordering instead of calling 2B-2 merely “next.”

• Round-2 verification at code checkpoint bfaaf2b: formatting and strict workspace Clippy; library 1,863 passed + 3 ignored default and 2,048 passed + 4 ignored CRDT; bottom-panel Stage 1 / 2A / 2B-1 / 2B-2 46 / 17 / 16 / 28; statusline 8 CRDT; semantic routing 2 CRDT; M4 121 passed + 3 ignored + 1 filtered; required GPU 202/202; isolated-config full workspace sweep; and git diff --check. The first workspace sweep had one GPU rendering failure in failures_and_display_math_render_as_source; that test had passed in the immediately preceding required-GPU run, passed alone, and the complete workspace rerun passed. Real-daemon and managed-attach cases were rerun outside the tool sandbox after its local-socket policy produced Operation not permitted; the authoritative reruns passed.

• Cross-machine recovery (fresh clone):

  git fetch githubsucks --prune
  git switch --track -c bottom-panel-stage2b2 githubsucks/bottom-panel-stage2b2
  git rev-parse HEAD
  

  #187 has landed, so githubsucks/main already contains this work and the branch is retained only for provenance. 2B-3 was cut from main.

• Stage 2B-1 MERGED as #184 (main @ 6bee09d, 2026-07-28; all twelve checks green on the reviewed head 5539b6e; two review rounds plus a gate-found follow-up). Branch githubsucks/bottom-panel-stage2b and worktree ../pmacs-bp-stage2b are retained and carry nothing unmerged. Durable facts — the schema-support-versus-advertisement split, the shared wire_grid boundary, authoritative Absent, and the two epochs — are in docs/agent-handoff.md §1 per rule 3, not here.

• What 2B-1 deliberately did not do, because 2B-2 and 2B-3 must not re-litigate it: no producer, no consumer, no capability change. panel_capable is still false for every semantic session, so the journey grade is unchanged and every shipped v20 client remains attachable.

• 2B-3 discharged the hard constraint 2B-1's review set: the compatibility-preserving activation is the frontend counter-offer described at the top of this lane, and the unsolicited Hello is unchanged. 2B-1's real-daemon acceptance for the v20 rejection point still passes untouched.

• Two review rounds, and what each cost. Round 1: PanelFrame needed an explicit buffer_id, the transport ratchet had to drive the real attach path rather than a detached codec assertion, and shared grid bounds needed one validator. Round 2: the server-first Hello made the advertised v20↔v21 compatibility one-way; COHERENCE.md and the handoff still named only v20 schema support; framing §9 named a nonexistent aggregate 2B suite instead of the three exact slice suites; and the panel plus copied-terminal "one byte over" fixtures were actually two bytes over.

• The full gate — not review — found two version-ladder omissions and one probe contradiction. The statusline and Vterm Stage 3 ladders still pinned v20 and rejected v21. Separately, Vterm Stage 3's headless probe exited its loop as soon as resize plus two nonuniform composites were observed, while its acceptance later required the PTY child's VTERMROW output in the final frame; the v20-compatible handshake made that scheduling race deterministic, so the report sampled a blank frame. The probe now waits for the exact child-output observation its acceptance asserts.

• The probe fix then leaked its own fixture, and that is the reusable lesson. The generic runner hard-coded the producer fixture's VTERMROW breadcrumb, so the CAT input fixture could satisfy every assertion yet never satisfy the loop exit — it waited out the 20-second safety deadline and passed on the deadline. Producer probes now name their own required frame text while input probes finish on the latched echo, and the report exposes completion_observed which both paths assert, so a deadline-driven pass cannot hide a stall again.

• Final verification on the reviewed head: cargo fmt --check; strict workspace Clippy; library 1,849 passed + 3 ignored default and 2,034 passed + 4 ignored CRDT; bottom-panel Stage 1 / 2A / 2B-1 46 / 17 / 16; folding Stage 2 48; GPU font 11; statusline 8 CRDT; m11_5 semantic 2 CRDT; GPU initial target and invocation 15 / 15 CRDT; the handshake consumers m5_5 / m5_7 / mode-system wiring 36 / 7 / 1 CRDT; Vterm Stages 1 / 2 / 3 10 / 6 / 9 CRDT, including the required real daemon + real PTY + real wgpu probe; M4 121 passed + 3 ignored + 1 filtered; required GPU 202/202; the isolated-config one-invocation full workspace sweep; and git diff --check.

  • Retained as classification rather than erased: the first required-GPU pass was 201/202 on a_fraction_draws_rule_pixels_between_its_operand_rows, a rendering test structurally outside a protocol-only diff, which passed immediately in single-threaded isolation and 202/202 on the mandatory complete rerun. Separately, library and Vterm attempts inside the restricted tool sandbox produced Operation not permitted failures in socket-based attach tests; the authoritative outside-sandbox reruns passed.

• Ordering for the rest of the arc is fixed: 2B-2 branches from landed main; 2B-3 branches only after 2B-2 lands; Stage 3 (the adopter default flip) last. Each slice starts fresh from landed main.

• Stage 2A MERGED as #177 (main @ 0a3fcd1, 2026-07-26, all twelve checks green at 8424172, three review rounds). Branch githubsucks/bottom-panel-stage2a and worktree ../pmacs-bp-stage2a are retained and carry nothing unmerged. Five commits: the classified census routing, the painter extraction + acceptance, the lane record, then the round-1, round-2 and round-3 review fixes. No protocol change; no behavior change for any frontend today — with panel_capable = false for semantic sessions, primary_document_window returns view.active in every existing configuration, so this is seam adoption that becomes load-bearing in 2B.

• Stage 2A verification on its merge result: cargo fmt --check clean; strict workspace Clippy clean; 1,832 default + 2,015 CRDT library tests; bottom_panel_stage2a_acceptance 17; bottom-panel Stage 1 46; statusline segments 8 CRDT; m11_5 semantic 2 CRDT; GPU initial target 14 CRDT; terminal config 12 CRDT; vterm Stage 1/2 10 / 6; folding Stage 2 48; M4 121; required GPU 202; git diff --check clean.

• Every routed producer is now pinned at a seam its production caller uses, and each pin was falsified by revert: #1 follow, #2 lazy CRDT upgrade, #3 CursorByte, #5 decorations, #7 Viewport (aligns without focusing), #8 Pointer (aligns and focuses), #9 the terminal-context gate, #12 statusline, #21 the publication filter, plus the focus-class negatives. #1/#3/#21 required extracting three named helpers, because their only production caller is dispatcher_loop, which no test can drive.

• Three lessons about the TESTS, not the code, all from review: (a) a structural test comparing the two authorities directly does not catch a misrouted consumer — only consumer-level assertions do; (b) a daemon-path test must register_session or the event is dropped at the uninstalled-session check before reaching the code under test; (c) a discriminating fixture must make the two routings DISAGREE — comparing two non-terminal buffers, or two windows with no selection, yields the same answer either way and proves nothing. Round 2 found four of my own pins vacuous by exactly these shapes, and round 3 found two more problems of the same family: a pin placed at a HELPER while production called it from a producer (reverting only the producer's call site left every test green), and a socket-pair assertion whose blocking read made a regression HANG instead of fail. Both now assert at the producer, with read timeouts on every read.

• Review round 1 closed: 4 P1 + 2 P2, all real. The P1s were a stale-Pointer focus steal (the failed-alignment arm returned the window, so #8's activation focused it before dispatch_pointer rejected the buffer), the missing A2A-2 two-context fan-out, a census suite that asserted the AUTHORITY rather than the CONSUMERS, and the missing main integration. Two of the new pins were themselves vacuous on the first attempt — the dispatcher test passed because an unregistered session is dropped at daemon.rs:1962 before reaching the aligner, and the painter test was a fixed-point check that survived deleting text_view.render. Both now fail under their own bite.

• vterm_stage3_acceptance::a37 is a pre-existing flake here, not a Stage 2A regression: measured 6/8 failures on the base commit and 7/8 on the branch in matched isolated samples. It needs a real daemon + real PTY + headless GPU and is documented load-sensitive. It also silently returns ok unless pmacs-gpu has been built, and is crdt-gated so CI never runs it at all.

• Two suites are dark without --features crdt: m11_5_semantic_acceptance reports 0 tests and gpu_initial_target_acceptance reports 1 in the default config. Both are semantic-census suites, so Stage 2A must be gated with the feature on or its most relevant coverage never executes.

• Stage 1 merged as #155 (main @ e745068, 2026-07-24, after two review rounds). No protocol change. Durable substrate facts live in docs/agent-handoff.md §1; the two round lessons are in §5.

• Landed-docs follow-up merged as #156 (main @ d152120, 2026-07-25).

• Stage 2 framing: docs/bottom-panel-stage2-framing.md revision 6 is on branch githubsucks/bottom-panel-stage2b (revision 5 is commit 56301ed there), worktree ../pmacs-bp-stage2b. Revisions 1–4 remain on githubsucks/bottom-panel-stage2-framing (head 4fbd47f, four framing commits, revision 4 at 49757e5). Round 1 closed 2 blocking + 3 high; round 2 closed 1 blocking + 2 high + 1 medium and decided both open items; round 3 closed 1 blocking + 1 high + 1 medium. No open items remain. Revision 5 adds no decision; it records the approved 2B-1/2B-2/2B-3 implementation split. Revision 6 corrects the server-first compatibility contract, durable protocol claims, exact acceptance-suite names, and limit + 1 fixture. The parent framing docs/bottom-panel-framing.md (rev 4) remains authoritative, including its acceptance criteria 37–55.

• Retained, carrying nothing unmerged: branch bottom-panel and worktree ../pmacs-bottom-panel.

• Stage 2 ships as four serial implementation slices, each landing before the next branches: 2A = classified §1.3 census routing + paint_frame per-window painter extraction (with the active-window auto-scroll preparation), no protocol change; 2B-1 = reserved protocol schema v21, with production advertisement held at v20, (InstanceMessage::PanelFrame plus FrontendEvent::{FrontendCellGeometry, PanelResizeRows, PanelPointer}, gated both directions, each extended enum byte-pinned on its own previous final variant); 2B-2 = daemon panel projection and epoch machine; 2B-3 = compatible v21 activation, the GPU band, and the negotiated panel_capable flip. Stage 3 is the adopter default flip.

• Correction — this entry previously mis-stated the census contract. It is not "route every consumer through primary_document_window". Q#BP14 classifies the 23 reads into four classes and routes only the Projection class that way; focus/input (#13–#15, #23), focus chrome and surface-routed (#16–#19), and focus/session (#20) keep their own authorities. Rerouting them would break remote-op validation and application, DispatchIdle, presence, focused search/menu/completion routing, and terminal bell ownership. The Stage 2 framing carries the full table.

• The GPU document bottom is three boundaries, not one. text_area_bottom (pmacs-gpu/src/main.rs:8490) is today status_band_top, geometry_capacity_bottom, and document_text_bottom at once. Once a band is installed they diverge: the status chrome must stay pixel-identical at the physical window bottom while document consumers move. A blanket rewrite of that helper moves both together and passes an "everything moved" assertion, so the Stage 2 criterion asserts both directions in one scenario. The census is 20 production sites (8 status-owned, 12 document-owned) + 1 definition + 8 test sites = 29 matches; the framing carries the per-site table. The three easiest to misclassify are document completion :6140, minibuffer candidates :7351, and edge scrolling :8561 — each with its own visible symptom.

• Folding Stage 3 and this arc's Stage 2 both touch the semantic projection. Whichever is framed second re-scouts the other's landed state.

Folding lane (Arc 6) — Stages 1 and 2 MERGED; Stage 3 (GPU) is next

Both shipped stages are on main; nothing in this arc is in flight. Stage 3 has no branch and no framing yet.

• Stage 1 (headless fold engine) merged as #142, Stage 2 (grid/daemon collapse) as #149 — both under "Closed since the last snapshot".
• Retained, carrying nothing unmerged: branches folding / folding-tui and worktrees ../pmacs-folding / ../pmacs-folding-tui. The framings docs/folding-framing.md (rev 5) and docs/folding-stage2-framing.md (rev 4) are the approved artifacts Stage 3 re-scouts against.
• Stage 3 (GPU) obligations, already named by the framings — the starting point for its own framing doc: GPU collapse at TUI parity; caret/hit-test fold-awareness; the BufferSnapshot fold-mirror clear (parent R2-4 — without it, empty-after-revert diff suppression leaves stale folds on the GPU, the same trap class as #120); CRDT-origin and GPU-optimistic interactive unfold (parent R2-3); and flipping FrontendView.fold_projection to true for semantic frontends, which Stage 2 deliberately left false (Q#FD21).

Resource-op delete guard implementation — PR #190 OPEN, review round 2 closed

• Portable branch: githubsucks/resource-op-delete-guard-impl, worktree ../pmacs-rd-impl. Implements the framing merged as #186 (docs/resource-op-delete-guard-framing.md, revision 5 plus its new §§9-10). Position against main, as pasted command output rather than a remembered constant — main moved while this lane was being written:

  $ git merge-base HEAD githubsucks/main
  b7bf2c664475c36b60cf7c0361ea75cd3c3b2315
  

  That is the whole durable fact, and it is deliberately the ONLY number pasted here. An ahead-count cannot be recorded in the file it counts: writing it is a commit, so the value is stale by one the instant it is written, and the previous attempt at this entry proved it — a pasted 4 0 read 5 0 at the pushed head. Run git rev-list --left-right --count HEAD...githubsucks/main when you need it; the merge-base above is what tells you whether the answer is still meaningful.

  Re-measure the merge-base too before relying on it. main moved twice while this lane's round-1 fixes were being written (#192 and #193), then twice more during round 2 (#188 and #194). This branch integrates through #194. main has branch protection now, so a stale base is not merely untidy: all 12 checks must pass on the merging head, and a conflicting PR builds no merge ref at all, so a green run from before the move reads as current when it is not.

• The framing's §8 branch plan is superseded and cannot be followed. It says "one PR — #186, which becomes the implementation PR", written when #186 was still open. #186 merged as framing-only, so the implementation got its own branch and PR. Nothing about the decisions changes; only the branch plan. Both the framing's header and its §8 now say so on their own pages.

• Layer 1 (the primitive) and Layer 2 (the applier + server-request boundary) are both complete and both pinned through their production paths. The Layer 2 gap the first commit named — criteria 11, 11a-11d, 12, 13, 15 having no production-path pin — is closed.

• Review round 1 found four defects; all four are fixed and all four are recorded in the framing's new §9, because two of them were corrections to that document, and a correction living only in a test comment is invisible to the next reader of the framing:

  • P1 §9.3 — the preflight broke ordered resource ops. Every delete was judged against the filesystem's initial state, so a valid create X -> delete X (or rename A -> B -> delete B) was refused with a fabricated NotFound about a path the batch was about to create. A regression this lane introduced. Decision: defer, do not simulate — a delete whose target is related by component-aware path containment to a path an earlier op creates, renames, or removes is left to the primitive. Q#RD3 already calls the check a filter, not a transaction. edit ops are deliberately excluded, so the buffer-and-filesystem half still fires early for untouched targets (criterion 11c depends on exactly that).
  • P1 §9.5 — the required production-boundary acceptances were missing. Landed: 11, 11a-11d, 12 (both directions), 13, 15.
  • P1 §9.4 — mid-batch failures were misreported as complete aborts. apply_workspace_edit now returns nil, message, applied_op_count, and ONE renderer serves both the status line and the server's failureReason. All three callers updated.
  • P2 §9.2 — non-recursive deletes inspected descendants. recursive is now a parameter of the shared query. The counterexample is an orphan: a modified buffer at tree/gone.rs whose file is already gone blocked a non-recursive delete of the now-empty tree/.

• Review round 2 found two more defects; both are fixed and recorded in the framing's new §10:

  • P1 §10.1 — dependency filtering compared raw path spellings. create dir/./x -> delete dir/x was wrongly preflight-refused even though both operations name the same lexical path. The comparison now runs both sides through the registry's existing lexical pmacs.path.canonicalize normalizer before component-aware containment. This is deliberately comparison-only: operation execution still receives the server's original path, and no filesystem/symlink canonicalization was added.
  • P1 §10.2 — a failing first plan item could mutate while reporting “nothing was mutated.” apply_workspace_edit now returns an execution_started fact in addition to the completed-item count. Only parse/plan/preflight failures claim that nothing changed. Once execution starts, the shared renderer conservatively says the failing operation may have changed state. Criteria 22a and 22b pin both forms: a multi-edit text item whose first edit lands before its second edit fails, and a resource rename that creates destination parents before the filesystem rename fails.

• delete_verdict is narrowed, and #171 inherits the narrowed version. Q#RD6's shared query is this lane's to own; descendant matching is now reserved for recursive deletes. Q#RD5's "inspect widely, mutate narrowly" is unchanged in substance — "widely" means the set the op can actually destroy.

• Criterion 3's stated bite: fixed by fixing the SETUP, not the doc. The framing says it fails against buffer-first ordering. Against the first shipped setup it did not (a directory target with no buffer bound to it), and §9.2's narrowing would then have left that setup with no bite at all. The buffer is now bound to the exact deleted path — a file opened, then replaced on disk by a non-empty directory, so a non-recursive remove_dir fails with ENOTEMPTY deterministically and under any uid. Both stated pre-images now bite, so the framing's wording needed no amendment after all.

• The fake is one parameterized mode, not eight. PMACS_FAKE_LSP_MODE=applyeditplan reads its whole WorkspaceEdit from PMACS_FAKE_LSP_EDIT_PLAN and publishes the client's response to PMACS_FAKE_LSP_APPLYEDIT_SINK (written .part-then-rename, so a polling reader never sees a partial record). Fail-closed: an unreadable plan sends no applyEdit and reports itself through the sink. pmacs_fake_lsp is a cargo BIN resolved through env!("CARGO_BIN_EXE_..."), so every CI leg builds it and a missing binary is a build failure — there is deliberately no skip-and-return-ok arm.

• Criterion 15's stub is hosted in m4_acceptance, and the gate list moved with it. lsp_dispatch_seams_acceptance is struck from the framing's §7 gate list AND its §8 touch table in the same edit, under §8's permitted simplification. It is still run as a gate, because builtin/runtime/lsp.lua changed.

• Acceptance: criteria 1-16, §9's 18, 19a-19c and 20, plus §10's 21 and 22a-22b, all in tests/m4_acceptance.rs and prefixed rd. 28 tests.

• Bite verification uses scripts/bite with the positive control it gained in #192, merged into this lane. The pre-image for the round-1 fixes is this lane's own first commit 1873be6, not main — those defects were introduced by it. Per-criterion results are in the commit message. The round-2 criteria 21, 22a, and 22b each pass on the round-2 code checkpoint cb7fe81 and produce a clean assertion failure against its pre-image c804dd5.

• Gates green at the round-2 tree: fmt; clippy -D warnings; --lib 1863; --lib --features crdt 2048; m4_acceptance 149 passed, 3 ignored, 1 filtered; lsp_dispatch_seams_acceptance 15; dired_acceptance 25 and autosave_acceptance 29 (the framing's watch items); required GPU 202; full isolated-config workspace sweep; git diff --check clean. The only warning in the non-Clippy CRDT build is the pre-existing unused_mut in src/daemon.rs; strict Clippy is clean.

• Recovery from a clean checkout: git fetch githubsucks && git worktree add ../pmacs-rd-impl -b resource-op-delete-guard-impl githubsucks/resource-op-delete-guard-impl.

dired Stage 2a — rename/delete reconciliation — PR #196 OPEN, review round 1 closed

• Portable branch: githubsucks/dired-stage2-impl, worktree ../pmacs-dired-s2. Implements Stage 2a only of the framing merged as #171 (docs/dired-stage2-framing.md rev 9, §5/§6/§10 — the substrate transaction, no dired surface). Position against main, as pasted command output rather than a remembered constant:

  $ git merge-base HEAD githubsucks/main
  e003b81cdd577140fc77330bd4578d3090696877
  

  That base is the #190 merge, and #190 matters here specifically: Stage 2a adopts its delete_verdict refusal rather than reinventing one, and lifts its walk query out into editor_core::buffers_bound_under so the guard and both reconciliation seams cannot disagree about which buffers an operation touches. Re-measure the merge-base before relying on it — main has branch protection, all 12 checks must pass on the merging head, and a conflicting PR builds no merge ref at all, so a green run from before a move reads as current when it is not. Re-measured after round 1: main had not moved, so no integration was needed — that is a reading of the tree, not a standing fact.

• What 2b and 2c still owe, stated so the split boundary is auditable. 2a ships no user-visible surface at all and no dired code: the dired_acceptance count is deliberately unchanged at 25, and a moved count there would mean it touched something it should not have. 2b owes the mark and operation layer (m u U t d x D R w M), pmacs.minibuffer.confirm plus its src/editor.rs load-sequence line, pmacs.killring.push, dired's own resource.renamed subscriber, and acceptance 1–22, 33, 39–41. 2c owes mkdir/copy/remove_dir_all, JobKind 12 → 15, dired.recursive-deletes, and acceptance 42–47.

• The split boundary has not moved since rev 9. It was re-checked against this tree: #188 (generated-buffer immutability Stage 1) did not convert dired's paint, so §3.1's coordination note is still an obligation of that lane rather than a collision with this one, and nothing in this diff touches builtin/runtime/dired.lua.

• Two m4 rows were re-pinned, and that is a behaviour change to a landed lane's assertions. rd9 and rd14 pinned #190's deliberate restraint on the apply_resource_op delete arm — descendants stay orphaned, only the first of two duplicate path-bound buffers is reconciled — and both doc comments gave the same reason: widening would have routed N buffers through remove_buffer_and_fire, phase 2 without phase 1, leaving up to N windows on removed ids. EditorCore::reconcile_delete composes both phases, so the constraint is discharged and the old assertions became the defect. Each row now asserts BOTH directions — reconciled away and no window holding a removed id — and each direction is bite-verified.

• One framing claim is wrong and is corrected at the test, not silently worked around. §5's G1 says a stale captured path "materializes a phantom" by reaching resolve_target_buffer's NotFound arm. It does not: pmacs.buffer.find_or_open calls crate::file_io::load_file directly and maps the error, so a missing path raises, and the NotFound arm belongs to resolve_target_buffer, which serves pmacs.window.display_file and the startup/daemon target rather than that binding. The defect is real and smaller: the pcall swallows the raise, so the user is stranded wherever the last applied op left them. Acceptance 34 is restructured to bite on that (its plan edits another file first, which is what makes the restore observable at all) and the correction is recorded in the test's own doc comment.

• Two bites were vacuous as the framing specified them, and both reasons are worth keeping. Item 28's rename row cannot pin the walk's containment rule: reconcile_rename calls Path::strip_prefix to rebuild a descendant's tail, and that is component-aware too, so a string-prefix walk is silently corrected a second time. The row moved to the delete side, where the walk's verdict IS the kill list. Item 30's composition-order assertion was a tautology: the LSP attach leaves diagnostic last in the stack, and moving the last element to the end is a no-op, so a remove-and-re-push was indistinguishable from an in-place mutation; the row now pushes one more overlay after it and asserts that precondition explicitly.

• 23 acceptance criteria are bite-verified by executed mutation, each labelled OK (assertion) — none merely OK (COMPILE), and none vacuous. Items 25, 27, 28, 29 (both directions), 30 (both mutations), 31, 31b (both gates), 31d (both halves), 34, 50 (both mutations), 51, 52, 53b, 54, 55, plus the two re-pinned m4 rows in three configurations.

• Review round 1 found four defects; all four are fixed, and all four were the same shape — a failure that left state wrong and told nobody. Worth keeping as one lesson rather than four bugs: every one of them was a pcall or a discarded return value, and each looked like defensive coding.

  • P1 — delete refusals were silent. reconcile_delete_and_fire returned kept_modified and refused and both production callers discarded them, so a last-buffer refusal or the asynchronous modified-buffer race left the file gone and the buffer still bound to it — and the next C-x C-s recreates the deleted file. Reporting moved inside the shared seam, for the same reason the reconciliation lives there: a caller that has to remember to report is a caller that will forget. Channel is EditorCore::status; not pmacs.error, which is defined only by a test stub, so a report there would have been the same silence.
  • P2 — the LSP subscribers swallowed their own reconciliation failures. Ignored pcalls made the callback return successfully, so the all-must-succeed logger had nothing to log. A shared failure sink now attributes each step and raises after the loop, because a fix that aborts on the first failure would leave every other attachment unreconciled — that wrong fix is itself a bite-verified mutation.
  • P2 — forget_uri left purged requests live in the client. It dropped pending_routes and pending_external but not the ids send_request puts in LspClient.pending, and recorded nothing in cancelled_rids, so a server that never replies leaked the entry and a late reply surfaced as a generic unrouted response. The per-rid work is now extracted from drain_cancelled_externals as abandon_request and reused rather than copied.
  • P2 — acceptance 35 was unpinned even after the G1 correction. With a plain delete the forbidden path fallback is unobservable: find_or_open raises out of load_file and the pcall swallows it, so both assertions passed with the fallback present. The plan now deletes the origin's file and recreates it, which gives the fallback something to open. The corrected G1 explanation also reached the production comments, which still repeated the false resolve_target_buffer::NotFound story — a correction that stops at the test comment has only half landed.

• One round-1 pin passed with its own bug restored, and the reason is reusable. Acceptance 53 asserted contains("only.txt") for the buffer-name attribution — but the status line opens with deleted only.txt:, the deleted path's basename, so stripping the attribution changed nothing the assertion could see. Both halves now assert the buffer's own name, which for a path-backed buffer is the full path and which only the attribution can produce. A pin written to close a review finding is exactly the kind that passes with the bug restored, and the detector was running the bite rather than reading the assertion.

• 31 bites now, all executed, every one labelled OK (assertion) — the original 23 plus 8 for round 1 (report call removed; refusal reason unattributed; kept-modified name dropped; subscriber failures swallowed; the wrong fix that aborts the loop; forget_uri skipping abandon_request; and the forbidden path fallback restored, which must fail acceptance 34 and 35 independently).

• Verification at this head, each gate run to its own file and its own exit code checked (never through a pipe): cargo fmt --check clean; cargo clippy --workspace --all-targets -- -D warnings clean; cargo test --lib 1,876 passed / 3 ignored; --lib --features crdt 2,061 / 4 ignored; the new resource_reconciliation_acceptance 25 default and 25 crdt; dired_acceptance 25 and 25 crdt, deliberately unmoved; the frozen additivity gate m8_1 10 / m8_2 15 / m8_3 32, all unchanged; m4_acceptance -- --skip basedpyright 149 passed / 3 ignored / 1 filtered; lsp_multi_root_acceptance 13; lsp_dispatch_seams_acceptance 15; typed_edit_chain_acceptance 13; journey_acceptance 24 (the ratchet floor, asserted as a count rather than a colour); gpu_invocation_acceptance 15 crdt — and that number is only real with pmacs and pmacs-gpu built first, which is the a37 trap in §5: the same command reported 12 failures before the build and 15 passes after, so a red run there is not evidence of a regression until the binaries exist; PMACS_REQUIRE_GPU=1 cargo test -p pmacs-gpu 202; isolated-XDG_CONFIG_HOME workspace sweep with --no-fail-fast 3,559 passed across 104 suites, 19 ignored, 0 failed; git diff --check clean. Every one of those was run as its own step with its own exit status checked — never cmd | tail inside an && chain, which returns tail's status and has masked a real failure in this repo before.

• Ownership, per the framing's own warning. §16 says 2a must not run concurrently with Journey Stage 1b, because 1b's LSP spawn-failure reporting lands in builtin/runtime/lsp.lua's attachment lifecycle and 1b's compile/binding half touches src/editor_core.rs — the same two files 2a rewrites, where the conflicts are semantic rather than textual so a clean git merge proves nothing. 1b must not be started while this PR is open. No other lane in flight touches them: #188 is dired.lua/buffer.rs generated-buffer writes, and the bottom-panel and CI lanes are elsewhere.

• Recovery from a clean checkout: git fetch githubsucks && git worktree add ../pmacs-dired-s2 -b dired-stage2-impl githubsucks/dired-stage2-impl.

Generated-buffer immutability framing lane — MERGED AS PR #188

• Portable branch: githubsucks/generated-buffer-immutability; worktree ../pmacs-generated-immutability. PR #188 landed on main as 27b1185 on 2026-07-29, after forking from githubsucks/main @ ad41cf1 and integrating through 5e186c7 — #189 (clean), then #186 and #171 (docs/active-work.md conflict), then #187 (the same file again, after it removed the two landed framing lanes), #192 at merge commit 76cfaac, and #193 (docs/active-work.md conflict again) after revision 7's first push. Revision 6 was reviewed at head 55c3061; revision 7 closes that round. The retained branch is provenance only. Framing only — docs/generated-buffer-immutability-framing.md, revision 7, plus this lane. No runtime code, no protocol change.
• APPROVED and merged after six review rounds (thirty-two findings, twenty-two P1, ten P2). Revision 7 is the governing contract.
• Stage 1 implementation is PR #191, open. The boundary is explicit and has already been needed twice: #188 owns the acceptance contract; #191 adopts criteria and may not restate, narrow, or reclassify them. Where an implementation finds a criterion impossible, the framing is revised and re-approved first. The selection-anchor clamp is Q#GB6's to specify and #191's to implement, and both must describe the same rule. Round 5 found #191 had locally restated Stage 1 criteria 5 and 7 while #188 still carried the originals — a divergence neither lane's gates can catch, because each is green against its own description.
• Round 5's two corrections that other lanes need:
  • Window::Selection::anchor is an unclamped byte position and it PANICS, not merely dangles. Window::region (src/window.rs:472-479) clamps neither endpoint; region_bytes (src/editor_core.rs:4184-4191) hands the result to Rope::slice, which asserts at src/rope.rs:145. Reproduced by #191: select 0..30, shrink the buffer to two bytes, copy. The fix is clamp-or-clear in both notify_buffer_edit (src/editor_core.rs:1836-1850, clamps nothing today) and rebuild_views_for (:1865-1882, clamps cursor and view_top but not selection) — and the rule is already in the tree for the terminal's own selection type at src/terminal/view.rs:715-721. A stale anchor also reaches the presence broadcast (src/presence.rs:122-123).
  • Buffer's on_edit broadcast stops at the first error in FOUR places, so a view later in attach order keeps pre-edit offsets: src/buffer.rs:1288 (apply_edit / apply_edit_skip_intercepts), :1250 (the no-op early-return arm), :1033 (apply_remote_crdt_op — the replica import path), and :1543 (broadcast_on_edit, i.e. undo and redo). Any lane relying on a buffer-attached view staying in step with the rope is affected.
• What it frames. The class-wide half of the set_generated_contents invariant that docs/agent-handoff.md §4 and COHERENCE.md §14 both record as unfinished: Buffer::undo gates on ensure_writable() (src/buffer.rs:1302) and never consults the intercept chain, so the add_intercept-plus-bypass_intercept idiom leaves the rope writable and every affected buffer emptiable. All five families were reproduced by execution at ad41cf1, not inferred; the transcripts are in the document's §0 and §2.
• Recommended primitive: Buffer::apply_generated_edit(op), exposed as a { generated = true } option on the existing Lua mutators, with set_generated_contents reimplemented as its whole-buffer wrapper. It is the only candidate in which the buffer is never observably unlocked. Revision 3 pins the transaction (framing §3.4): its own run_buffer_edit arm — not the bypass arm, which calls begin_edit, which calls ensure_writable first (src/buffer.rs:725) and would refuse every generated write to a locked buffer — one &mut Buffer method with every exit named. Revision 4 replaces revision 3's cleanup predicate. Cleanup is driven by an explicit five-variant GeneratedOutcome reported by the apply, not inferred from revision. Inferring it was wrong three ways: a successful no-op (src/buffer.rs:1245-1253 returns Ok without bumping revision) kept history the contract forbids; a CRDT mid-transaction failure happens upstream of revision entirely (:1140-1163), so it was neither cleaned nor detected; and the unconditional relock locked a fresh buffer that was never successfully written. NoOp clears, Rejected restores the entry lock state, Diverged clears nothing and surfaces. Revision 5 keeps the five outcomes but preserves the Edit in AppliedThenFailed { edit, error }: the borrow-free Lua finisher fans it out to window caches and replica mirrors before returning the error. Collapsing to Result inside Buffer was too early. Revision 6 replaces the delete→insert enumeration with a crdt_mutated flag, because export_updates_since can fail after every successful CRDT op shape, and changes all four Buffer broadcast loops to continue-and-retain-first-error. Revision 7 installs quarantine at common divergence detection before generated or ordinary callers map the outcome. The monotonic poison blocks all three CRDT snapshot exporters, daemon-origin queueing and every later owner-generated write; criteria 16c and 16d pin the containment boundary. Repair remains deferred.
• Two stages, two PRs. Stage 1 — listview ownership fix plus its identity-routing fix in the same PR, dired and listview adopting the shipped primitive, the window-coordinate clamp, and the fold decision. Stage 2 — the new primitive, compile's nine write sites, the search panel's four, compile/search ownership + routing, the path-backed refusal plus mark_clean, and the terminal-only identity_protected guard. No Lua unlock ships.
• Nine facts this lane landed for other lanes:
  • bypass_intercept is the wrong inventory key. It misses *buffer-list*, *help* and *workers*, which are generated with plain writes and no intercept at all. docs/agent-handoff.md §4's four-row table inherits that blind spot — and undercounts by one: src/help.rs:354 replace_help_buffer is a fifth writer mechanism (own find-or-create, Buffer::apply_edit, own mark_clean) writing the same *help* buffer as default.lua:1239, which does not mark clean. Two owners, one buffer, two copies of the name constant across the FFI boundary.
  • COHERENCE.md §14's listview consumer list was wrong and is now FIXED — PR #189 (main @ 7586905) landed exactly the correction this lane measured. Nothing owed. Recorded so it is not re-asserted.
  • Three writers adopt any buffer sharing their name — listview.lua:95, compile.lua:263, default.lua:861-868 — against a rule the tree already states at terminal.lua:300-305 and implements at dired.lua:476-504. Measured: a foreign *references* is clobbered and left permanently un-editable, and a pmacs.compile.run that raises on validation still leaves a foreign *compilation* un-editable. Today M-x buffer.undo — this arc's bug — is the only recovery, so the arc must not lock these buffers before fixing ownership.
  • Disambiguating a name breaks the sites that read one. Census in framing §2.10: 19 units across 14 grep lines, two genuinely broken. listview.lua:44's panels[d.name] (written under the requested name at :97, read under the actual name) has four consumers, and the fourth — listview.open:118-123's never-capture-a-panel guard — fails inverted and silently, capturing a panel as its own q target. compile.lua:216's is_generated_buffer has two. compile.lua's slots is not affected: keyed by a module constant at both ends, with slot_for_buffer id-based.
  • read_only is one boolean serving THREE policies (framing §2.11): the generated lock; terminal identity (src/terminal/session.rs:305); and, as a reader, src/lua_bindings/fold.rs:313's "is this a document buffer" test, pinned by tests/folding_acceptance.rs:570. Consequence for any lane: locking a buffer silently disables pmacs.fold.fold on it, with the status fold rejected: not a document buffer.
  • The SHIPPED set_generated_contents can overwrite a live terminal identity buffer. It does self.read_only = false unconditionally (src/buffer.rs:546), so it lifts a lock it did not install, writes, and re-locks. Present on main, untested, unframed anywhere before revision 4. Refused in Stage 2 by the identity_protected field — an intrinsic flag marked once by a crate-private monotonic mark_identity_protected() in TerminalSession::open, never written by set_read_only. Revision 3 tried to infer this from the lock's provenance instead; that broke the lift-and-restore idiom at tests/terminal_copy_mode_acceptance.rs:578-584, and the general lesson is that a derived fact must be maintained by every mutation of what it derives from — and set_read_only is pub.
  • acc16e is crdt-gated and is the only shipped consumer of the lift-and-restore idiom. cargo test --test terminal_copy_mode_acceptance without --features crdt never compiles it, so a green run of that suite proves nothing about the seam. Any lane touching read_only semantics must run it with the feature and confirm acc16e is in the count.
  • identity_protected is not generated-lock provenance. Revision 4 tried to use “not a terminal identity buffer” as proof that the generated primitive installed the lock; it is not. Revision 5 therefore removes pmacs.buffer.unlock_generated from the arc entirely. Wdired's future generated→editable transition remains dired Stage 3 work and must be owner-specific or use the eventual lock-policy enum.
  • The CRDT Replace mid-transaction divergence is real and unowned. crdt.delete then crdt.insert (src/buffer.rs:1140-1163); if the first succeeds and the second fails, the code's own comment says "the CRDT is mid-transaction ... This is an invariant violation." It reaches apply_edit and apply_edit_skip_intercepts today and is reported as an ordinary CrdtRejected, so nothing distinguishes it. This lane names and contains it; repair is deferred and unowned. Revision 6 makes classification total with a crdt_mutated flag: any later error, including export_updates_since after Insert, Delete or Replace, is Diverged. Revision 7 installs a common monotonic poison before either generated or ordinary API mapping, blocks initial_target_snapshot, send_buffer_snapshots, export_buffer_snapshot and queue_daemon_origin_crdt_op, and rejects every later owner-generated write with CrdtQuarantined. Criteria 16c and 16d fault-inject the complete boundary under cargo test --lib --features crdt; there is no public fault-injection API and no four-variant fallback.
• Overlap warning. Stage 2 touches src/lua_bindings/mod.rs's buffer mutator bindings and src/buffer.rs. Do not run it concurrently with the apply_resource_op lane or the bottom-panel 2B work without assigning those files to one lane first. The framing itself touches neither.
• Cross-lane, settled, not re-decided here. #186 owns the urgent pre-filesystem refusal for synchronous apply_resource_op; #171 later owns full post-delete lifecycle reconciliation, including the async race where a buffer becomes modified after dired dispatch. #171's Q#DR25 is deferred INTO this lane — confirmed against #171 revision 7 (fd7ae37), which states that dired's listing becoming immutable is "owned by the generated-buffer-immutability lane" and that "Stage 2 does not implement it, does not gate on it, and carries no acceptance for it." This lane's Stage 1 claims that work. Neither ordering conflicts: #171 Stage 2b changes paint's callers, this lane changes paint itself. Revisions 1 and 2 of this framing never mentioned Q#DR25 at all; revision 3 §9b records it.
• Re-measured at ad41cf1 while scouting: 276 CRDT-dark tests (3,251 vs 3,527), by cargo test --all-targets --no-default-features --features lua54[,crdt] -- --list | grep -c ': test$'. Recorded here because the section above asks for exactly that and warns against quoting a stale figure; it does not replace that section's per-target census, which was not re-derived.

Test-improvement arc, lane 2 — silent-skip arming

• Portable branch: githubsucks/silent-skip-arming, worktree ../pmacs-skiparm. Implements TEST_IMPROVEMENT.md §1.2 and §5.4.

• Base, measured at write time rather than quoted:

  $ git log --oneline -1 githubsucks/main
  5e186c7 Merge pull request #193 from levineuwirth/test-improvement-audit
  

  The previous revision of this entry said "base measured at write time, pasted below" and then pasted nothing: the script meant to substitute it reported success and silently matched no text, and the claim was not re-read. Recorded because it is the same defect this ledger keeps catching one level up — asserting a measurement is not making one, and a tool reporting success is not the measurement either.

• Recovery from a clean checkout: git fetch githubsucks && git worktree add ../pmacs-skiparm -b silent-skip-arming githubsucks/silent-skip-arming.

• The defect: let Ok(_) = which_binary(x) else { eprintln!(..); return; } reports GREEN when the tool is absent, and CI installed none of the tools. A block of real-language-server and multi-shell tests had therefore never once executed their bodies in CI while reporting success. A suite that cannot distinguish "passed" from "never ran" is worse than a missing one, because it reads as coverage.

• The fix is the project's own pattern. PMACS_REQUIRE_* already makes a missing GPU fatal for vterm_stage3_acceptance; this adds PMACS_REQUIRE_LSP, PMACS_REQUIRE_SHELLS and PMACS_REQUIRE_LUA, plus the CI step that installs the tools. Per-tool variables, not one blanket flag, so a tool that must stay unarmed keeps its decision visible at the call site.

• basedpyright is deliberately NOT installed and NOT armed. Its test has no timeout and hangs forever — root cause is the non-interruptible reader-thread join in RuntimeHandles::drop, already a named deferral in src/process.rs. The test job has no timeout-minutes either. Arming it today would trade a vacuous green for a six-hour hang across four legs. PMACS_REQUIRE_PYRIGHT exists and is never set, so the flip is one line once lane 4 (the hang) and lane 3 (timeouts) land. Do not arm it before both.

• A trap found while writing the workflow, not after: the natural Actions idiom ${{ runner.os == 'Linux' && '1' || '' }} sets the variable to the EMPTY STRING elsewhere, and var_os().is_some() is true for Some(""). That would have armed the guard on exactly the runners with no tools installed. The helper therefore treats empty as unset. PMACS_REQUIRE_GPU has the same latent shape and is safe only because it is set literally.

• Verified by execution in all three states, on a tool genuinely absent from this machine (vscode-json-language-server): unset -> skips green; armed -> hard failure naming the CI step; empty string -> skips green. The armed failure is the bite, and on main it cannot occur because no guard exists.

• The tests pass when they actually run — which was the open question, since none of them had. Armed locally: 11 m6_5 + 8 m6_8 REPL tests green, and all six real-LSP tests (clangd x2, gopls x2, rust-analyzer x2) green individually.

• rust-analyzer is installed in the Linux-gated step, not via the toolchain action's components:. The first revision put it there, which applies to every matrix leg — and presence, not PMACS_REQUIRE_LSP, is what decides whether a gated test body runs. That would have executed the two rust-analyzer tests on macOS for the first time ever, on the legs that are simultaneously the CI critical path and the documented flake surface, while this entry claimed Linux only. The variables not being set there would only have meant absence was tolerated; it would not have kept the tests skipped. Text and workflow now agree.

• Tool versions are pinned (gopls@v0.16.2, vscode-langservers-extracted@4.10.0, yaml-language-server@1.15.0). @latest and bare npm install -g make CI drift with upstream releases, so a bad publish breaks CI with no commit here to bisect against. Caching the built gopls on the pinned version is a follow-up, not done here.

• §1.2 is NOT fully closed by this lane. The guards arm the entry skip only. tests/m4_acceptance.rs's mid-test rust-analyzer bail ("workspace likely still indexing; skipping") survives, so even armed, that test's only assertion can still vanish under load — precisely when a regression would show. Mid-test skips are their own shape and want their own pass.

• Not this lane's to fix, recorded so it is not mistaken for oversight: the generated-buffer immutability lane above still reads "PR #188 OPEN, PROPOSED" and #188 has merged. Rule 4 forbids relabelling it and permits removal only once its durable facts reach docs/agent-handoff.md, which #188 did not touch — it changed the framing and this ledger only. So the absorption is genuinely owed, and the natural carrier is the arc's own next PR (#191, Stage 1), not a testing lane reaching across into someone else's arc.

• Follow-up owed after this merges: delete githubsucks/handoff-2026-07-20. Removing the documentation lane removes the only pointer to that branch, so nothing will otherwise remind anyone it still exists on the remote.

• Linux only for now, deliberately: macOS needs the brew equivalents and roughly doubles install cost on the slowest matrix leg. The variables stay unset there, so those tests skip cleanly.

• Also removes the documentation lane, whose disposition the ledger left undecided pending confirmation that its branch carried nothing unique. Confirmed by measurement: githubsucks/handoff-2026-07-20 is 1 ahead, 365 behind, and its entire unique diff is four doc files at 42 insertions against 88 deletions — merging it would revert current documentation. The section said "whoever confirms the branch carries nothing unique removes the section"; this is that.

Test-improvement arc, lane 3a — CI timeouts and concurrency

• Portable branch: githubsucks/ci-timeouts-concurrency, worktree ../pmacs-ci3. Workflow only — no product code, no tests changed.

• Base, measured at write time:

  $ git log --oneline -1 githubsucks/main
  b7bf2c6 Merge pull request #194 from levineuwirth/silent-skip-arming
  

• Ships the three cheap, deterministic items of TEST_IMPROVEMENT.md §5-6. The larger ones — nextest (§6.3), the serial/parallel split (§6.2), the parallel canary leg (§5.6), the nightly cron (§5.5), and the macOS matrix trim (§6.4) — are deliberately not here: each changes what CI certifies or how it runs, and each wants its own decision rather than riding a timeout patch.

• timeout-minutes on every job (§5.2). Measured before changing: 7 of 8 jobs had none and inherited GitHub's 360-minute default; only m6-perf-gates had one (15). A single hung test therefore burnt six hours, times four on the test matrix. This is the gate that must land before PMACS_REQUIRE_PYRIGHT can ever be set — lane 2 left basedpyright unarmed precisely because this did not exist.

• The ceilings are 25, and 35 for the test job — anchored on observed execution, corrected in review. Revision 1 cited "~14.6 min, ample headroom", which was one reading quoted as a property. Re-measured over two windows: 17 min max over 25 runs and 15.8 min over 12, both macOS/luajit; every other job under 4 min. Against 17, a flat 25 is ~1.5x, not "ample".

  • timeout-minutes counts execution, not queue — a 33-minute wall-clock run in that window executed its longest job in 17 — so no run in observed history would have been killed by either value.
  • The real exposure is what the window does not contain: a cold cache. A stable-toolchain bump invalidates Swatinem's key on every leg simultaneously, and a cold macOS debug build plus suite is the plausible way a healthy run overruns. It would present as four legs timing out at once, the day after a Rust release.
  • So the test job takes 35 (~2x its observed max) and the rest keep 25 (~6x theirs), and the diagnosis is written into the workflow before the event: simultaneous four-leg timeouts after a toolchain release are a cold cache, not a hang; a single leg timing out beside passing siblings is the hang case.

• concurrency with cancel-in-progress (§6.1), scoped to pull requests. github.event.pull_request.number is empty on a push to main, so the fallback keys those by SHA and no main run can cancel another — cancelling one would leave the branch-protection record ambiguous about a commit that already landed.

• -p pmacs-protocol clippy (§5.7). Verified passing locally before proposing it, so adding it cannot turn CI red on arrival. The root-package clippy never covered it: the workspace default member is only pmacs.

• §5.1 branch protection is DONE, not deferred — it belongs in neither this lane's shipped list nor its deferrals, and review was right that its absence from both was an omission. It was enabled earlier in this session; verified against the API at review time:

  $ gh api repos/levineuwirth/pmacs/branches/main/protection
  {"enforce_admins":false,"force_push":false,"required_checks":12,"strict":false}
  

  All 12 checks required; strict off deliberately, so a PR need not rebase every time main moves (this repository's ledger contention makes strict expensive); enforce_admins off so the user retains an override. This matters to the concurrency comment, which justifies exempting main pushes by appeal to "the branch-protection record" — that record now exists, so the justification is real rather than aspirational.

• Required status checks are NAME-COUPLED to job names, and this lane's own deferrals will break them. A required context that no longer exists does not fail — it leaves every PR pinned on "Expected — waiting for status", indefinitely, which is main becoming unmergeable by policy rather than by a red run. Three deferrals above change job names or the matrix: the macOS trim (§6.4) removes two contexts outright, and nextest (§6.3) or the serial/parallel split (§6.2) rename or add them. Rule: any job rename, removal, or matrix change updates the branch-protection required-checks list in the same motion. Recorded here because this is the entry that both enabled protection and named the lanes that will invalidate it.

• Recovery from a clean checkout: git fetch githubsucks && git worktree add ../pmacs-ci3 -b ci-timeouts-concurrency githubsucks/ci-timeouts-concurrency.

Test-improvement arc, lane 4 — process teardown stdin deadlock

• Portable branch: githubsucks/process-teardown-stdin-deadlock, worktree ../pmacs-hang. Implements docs/process-teardown-stdin-deadlock-framing.md (rev 3: one review round, then a CI round that falsified the reproduction).

• Base, measured rather than quoted:

  $ git log --oneline -1 githubsucks/main
  e003b81 Merge pull request #190 from levineuwirth/resource-op-delete-guard-impl
  

• Recovery from a clean checkout: git fetch githubsucks && git worktree add ../pmacs-hang -b process-teardown-stdin-deadlock githubsucks/process-teardown-stdin-deadlock.

• The defect: RuntimeHandles::drop joined its reader threads in the Drop body, which runs before any field drops. The ChildStdin sink lives in the stdin field, so it could only be released after the join returned — and the join waited on readers blocked in read() on pipes whose write ends the child still held, because the child never got the stdin EOF that would have made it exit. A closed cycle inside one function; teardown hung forever.

• This is the root cause of the m4_5_basedpyright hang that has parked --workspace sweeps (once for 2h26m) and forced -- --skip basedpyright into every gate recipe. The handoff's §3 claim that the desktop's binary was broken is retired by this PR: the binary was fine. basedpyright-langserver is a uv console script that runs bundled node via subprocess.run and waits; at teardown shutdown() SIGTERMs the recorded pid (the wrapper), which dies without forwarding, and that orphans node to PPid: 1 holding the pipes. A direct binary like clangd is a genuine child whose pipes close on reap. That is the whole of the "intermittent" story.

• Corrected in review round 2: rev 1–3 said the wrapper "spawns node and exits". Wrong — and refutable from evidence already in hand, since the initialize handshake succeeds, which a wrapper that exited at spawn could not have done. The PPid: 1 observation was taken after shutdown() had killed the wrapper. We create the orphan. The fix is unaffected; the parked follow-up changes from "tolerate self-orphaning servers" to "stop orphaning them" (signal the group).

• Diagnosis method, because reproduce-first was the instruction: gdb thread stacks plus /proc fd forensics on a live wedged process, both pipe ends identified in both processes, reproduced 5/5. Three earlier reproductions were vacuous — see the handoff §5 lesson; the shipped test carries two positive controls because of it.

• Verification (each gate its own step, real exit status, no cmd | tail): fmt 0; git diff --check 0; clippy 0; --lib 1864 passed; --lib --features crdt 2049 passed; m4_acceptance without the skip 150 passed in 2.66s with the basedpyright test ok; the eleven PTY/REPL/worker/panel suites of the framing's Bet 2 all 0 (144 tests); PMACS_REQUIRE_GPU=1 -p pmacs-gpu 202 passed. Bite verified by revert: ok in 2.03s with the fix, FAILED on timeout at 10.00s without it, both controls passing first.

• CI round 1 falsified the reproduction, and the control is what caught it. Three Test legs failed on 9b1cf3d's predecessor: the synthetic child used sh -c 'cat <&0 & exit 0', and <&0 does not defeat the /dev/null rule it was chosen for — the rule applies before explicit redirections, so fd 0 is already /dev/null and the redirect duplicates it onto itself. bash skips the default when a stdin redirect is present; dash, which is Ubuntu's and CI's /bin/sh, does not. Local probing through /bin/sh could not see it. Now setsid --fork cat, with no shell at all. Lesson recorded in the handoff §5: never probe shell behaviour through /bin/sh — name the implementation.

• acc28 on macos/lua54 was a flake, established not assumed. bottom_panel_stage1_acceptance::acc28 failed once on that leg; rerunning the same job on the identical head passed, and the suite is 46/46 locally. It is now in Bet 2's falsifier list — its absence from rev 1 was a real gap, since it drives real child input through a PTY in a panel and this PR changes PTY-mode teardown ordering.

• Not fixed here, parked in the framing §5: cancellable non-group read (covers a child that ignores EOF, and one that stops draining while write_all is blocked); the orphaned-server leak — post-fix the server exits by cooperation, not enforcement.

• CLAUDE.md's --skip basedpyright entry is deliberately untouched. Dropping it is a separate proposal owed evidence of repeated green. The timeout precondition is already satisfied — #195 (this PR's base) gave every job a timeout-minutes — so the only remaining reason PMACS_REQUIRE_PYRIGHT stays unarmed is that CI does not install basedpyright at all; arming it would fail rather than test anything.

• Adds PMACS_REQUIRE_SETSID, armed on Linux. The teardown test's fixture needs setsid --fork, which is util-linux rather than coreutils, so it skips when absent (the standard --lib gate must not hard-fail a minimal container on an undeclared tool) and the variable makes that skip fatal where the tool is guaranteed. Both arms verified against a PATH with setsid genuinely removed: unarmed skips, armed FAILS. README's test-dependency list declares it.

Parked lane: kill-ring browser + persistence

• Portable branch: githubsucks/kill-ring-browser
• Parked framing head: 503c489
• State: framing only, revision 2; no implementation and no PR.
• Status: explicitly parked by the user on 2026-07-20.
• Its original scout was based on 0efb5cd. The preserved framing marks this ground truth stale and requires a complete re-scout against the then-current githubsucks/main before implementation.
• Compile-mode has merged since the original scout, so old “compile-mode in flight” keybinding/touch-set assumptions are not authoritative.

Recovery worktree, only when the user un-parks it:

git worktree add --track \
  -b kill-ring-browser \
  ../pmacs-kill-ring-browser \
  githubsucks/kill-ring-browser

Closed since the last snapshot

• Terminal configuration + copy mode arc — BOTH STAGES MERGED, lane removed. Stage 1 #173 (main @ cf54270, one review round) and Stage 2 #178 (main @ fe8b8ba, four review rounds, twelve checks green on head 1b44c69 — verified by head_sha, not by the check summary), both 2026-07-26, both with no protocol change. Approved framing: docs/terminal-config-and-copy-mode-framing.md rev 4, committed as the first commit of Stage 1's branch; its Q#TC6a carries a superseded-in-part box rather than a silent rewrite. Durable facts moved to docs/agent-handoff.md §1 (the arc bullet) and §4 (the set_generated_contents invariant) per rule 3 below, and to COHERENCE.md §14. Stage 2 ships eight of nine criteria and the missing one is named — criterion 17 needs a real GPU frontend, so it waits on the a37 footing; the handoff records what it must assert. Branches githubsucks/terminal-config and githubsucks/terminal-copy-mode with worktrees ../pmacs-terminal-config and ../pmacs-terminal-copy-mode are retained. The gate-run flake found while gating #178 moved to the CI crdt-coverage lane above, which owns its discrimination.

• Dired Stage 1 (the directory view) — MERGED as #165 (main @ c8ec8f3, 2026-07-25, after one review round). pmacs has a directory surface: C-x d / C-x C-j, one read-only buffer per directory named *dired:<canonical path>*, a dired major mode carrying RET/f, ^, n/p, g, q, s. No wire change (v20). The Rust is two things — a per-entry-tolerant read_dir (Q#DR6), which had to be Rust because read_dir_blocking fails a whole listing on any of five per-entry conditions and a tolerant wrapper cannot be written in Lua at all, and normalize_buffer_path going pub as pmacs.path.canonicalize (Q#DR2's preferred end state, so no Lua mirror exists and Stage 2 owes no mirror removal). The frozen m8_1/m8_2/m8_3 counts are unchanged, which is the additivity gate. 15 claims bite-verified; one came back VACUOUS (acceptance 3c cannot pin descent routing — dired holds focus in its own panel, so dedication is the only discriminator) and is documented at the assertion rather than relabelled. Its branch (dired-stage1) and worktree (../pmacs-dired-stage1) are done; the abandoned dired branch (ffdd642, ../pmacs-dired-arc) was superseded by a fresh cut and carries nothing unmerged. Stage 2 (marks and operations) and Stage 3 (wdired) each still need their own framing, and the frozen fixture shrinks after Stage 3. Durable substrate facts and both new ops lessons live in docs/agent-handoff.md §§1/5; the implementation notes are docs/dired-framing.md §0, S1-1…S1-12. Two named forward items for Stage 2: apply_resource_op's rename rebind is exact-PathBuf-equality, first-match-only, looked up with the raw path while stored paths are normalized — so a directory rename strands every buffer under it, and pmacs.fs.rename has zero production callers, so it can be fixed at the primitive; and Q#DR5's seam is the main-thread drain AsyncRuntime::tick, not _take_result, where rename settles as an undifferentiated ReplyKind::FsUnit and so must be keyed on JobKind::FsRename.

• GPU terminal input (the double terminal-layout sync) — MERGED as #166 (main @ b889873, 2026-07-25, one review round, all twelve checks green after a macOS PTY-timing rerun). The dispatcher applied both terminal-layout syncs to every attached frontend; a semantic session satisfies both conditions, so its PTY was resized twice per tick forever and the child took a SIGWINCH storm that made a GPU terminal untypable while output still flowed. sync_terminal_layout is now split into a frontend-kind-neutral half (panel reconcile + controller liveness) and a grid-only geometry half, with the loop body extracted to sync_terminal_layouts_for_tick so the exclusivity is structural. No protocol change (v20). Durable lessons are in docs/agent-handoff.md §5; the framing (docs/gpu-terminal-input-framing.md rev 2) carries three falsified hypotheses, the two-pre-image bite matrix, and two named out-of-scope items (Q#GT5 interactive-shell echo on a raw PTY, which reproduces in-process and so is not the GUI/TUI asymmetry; and a geometry change appearing to clear the visible screen, which reproduces pre-fix). Branch gpu-terminal-input and worktree ../pmacs-gui-term-input retained. Its landed-doc pair MERGED as #168 (main @ 1b6a084, 2026-07-26): #166 recorded as landed, the CI crdt-coverage gap measured (264 tests dark workspace-wide, 177 in the library — a reading taken at 1b6a084 and kept here only as history. The CI crdt-coverage lane above is the authority for the live figure; do not quote this one forward), the vterm audit corrected — "only 3 of 9 acceptances drive a real daemon" was optimistic; without the frontend binary the honest number is 2 — and the a37 findings folded into the coverage lane.

• Inline-math slice — MERGED as #158 (main @ 5aa9044, 2026-07-25). Detect → parse → layout → draw for $…$, entirely inside pmacs-gpu, no protocol change. Verified by the user's manual pass on a real paper after the landing. What is worth carrying forward:

  • The v0 subset is 34 Greek symbols, sub/superscript, and \frac. An unsupported command fails the whole span back to source, so on a real document most inline spans still show LaTeX. Widening the symbol map is the highest-value next increment — ahead of display math, which is also deferred.
  • A stale frontend binary is invisible from the source tree. The slice lives only in pmacs-gpu, so after it merged the feature was absent until cargo build --release -p pmacs-gpu and a client restart; the daemon needs neither. Diagnose with strings on the binary (Latin Modern Math, MathBox) rather than by re-reading the checkout, which was already current.
  • Main was integrated three times in one day (8c86d34, 46a1b8f, b889873), merged not rebased to preserve review anchors. Two conflicts, both this ledger and nothing else. The dangerous case was the one that did NOT conflict: #166 auto-merged into pmacs-gpu/src/main.rs, the file this lane rewrites, because the two edits sat in different regions of it. Decide whether to integrate from the shared-file set, never from whether git complained.
  • Integration was proved by test-count reconciliation, not by a green run: predict what the other side adds, then check the deltas. GPU 199→202 matched e547a90's 3; later lib 1,826→1,829 and CRDT 2,003→2,006 matched #166's 3, with GPU unchanged because #166 adds none. Suites 91→92 was #161's new binary.
  • Why the branch had no CI for a day: a conflicting PR builds no merge ref, so no pull_request run is created. The ledger previously recorded this cause as unidentified; it is not. Check mergeable and confirm a run exists for the current head SHA.
  • m4_5_basedpyright has no timeout and hangs forever, parking a --workspace sweep (observed 2h26m at 38 of 92 suites). It is intermittent, so an earlier clean sweep proves nothing. Sweep with cargo test --workspace --no-fail-fast -- --skip basedpyright and judge progress by whether the suite count advances.
  • Named v0 approximations: the peer-caret half of acceptance 14 is pinned at the mapping level, not pixels; a soft-wrapped spacer draws its box whole at the first run's origin; the fit budget reads the bundled code face even under a custom set_font family.

• Bottom panel Stage 1 — MERGED as #155 (main @ e745068, 2026-07-24, after two review rounds). Window placement, window parameters, TUI side windows, the divider, and the adopter display opt-in, with no protocol change. Both rounds found the same class of defect and are worth keeping:

  • Round 1: the Q#BP6 side-window split guard had no production caller — C-x 2 still reached plain split_active — and survived because the acceptance test called the core method directly.
  • Round 2: Q#BP7's terminal growth re-arm had never been implemented, and the assertion meant to pin it (at_bottom) is a geometric readout that a still-anchored view satisfies; the anchor assertions beside it compared "" with "" because the PTY fixture emitted LF-only output.
  • Post-round-2 self-review, caught by CI going red on all four Test jobs: resolving pmacs.window.buffer()'s no-argument arm through the acting frontend made a total function partial, and six runtime modules silently dropped operations (kill_ring_acceptance 30/30 → 25/5). Fixed in 9110f9f before merge.
  • Gating fact found on the way: the workspace sweep must run with an isolated XDG_CONFIG_HOME, because the real user init.lua installs a local package and the losing race leaks a status message into painted-frame comparisons. There is also a latent pre-existing main bug in the buffer CRDT undo path, unrelated to this arc.
  • compile_mode_acceptance is load-sensitive under default parallelism (~1 run in 3, a different test each time); verified pre-existing by swapping in main's compile.lua. It is 67/67 at --test-threads=1.

• GPU initial target — MERGED as #148 (main @ 0dd16a5, 2026-07-24, after two review rounds). pmacs --gpu [--socket …] FILE opens a target before the GPU window appears. Protocol bumped 19 → 20: a semantic-session SessionBootstrapRequest after AttachRequest, plus an appended InstanceMessage::InitialTargetResult pre-window readiness barrier; v6–v19 wire encodings are unchanged. Root owns launcher tilde/cwd resolution and exact raw-byte path transport; the daemon resolves/dedups/loads the target and runs load/switch hooks inside one dispatcher transaction, then publishes CRDT-upgraded targets to existing grid replicas (gated on upgraded_to_crdt, independent of the load/create outcome, so a dedup onto a hidden not-yet-backed buffer still reaches pre-attached replicas — round 2 finding). Semantic replicas receive a publication only when displaying that buffer, so a second target launch cannot switch an existing GPU window. Round 2 also closed a failure-containment gap: every dispatcher-side bootstrap failure now shuts down the socket (a dropped write-half clone does not close a shared FD), and the dispatcher drops any event from a session that was never installed, rather than reaching absent render/size state. Integrated cleanly with Folding Stage 2 (#149): fold projection at attach is selected from the same negotiated semantic_render bit the target bootstrap uses. Its lane, worktree (../pmacs-gpu-initial-target), and branch (gpu-initial-target) are done; the -framing branch is kept. Durable substrate facts and both review-round lessons live in docs/agent-handoff.md §§1/5 and docs/gpu-initial-target-framing.md rev 3.

• Folding Stage 2 (grid/daemon collapse) — MERGED as #149 (main @ 6ed4fe9, 2026-07-24, after five review rounds). The grid TUI now renders collapses. Spine (Q#FD12): src/fold_view.rs's VisibleLineMap, derived from the fold store plus a window's line offsets and never stored, threaded as Option<&'a VisibleLineMap> on a lifetime-bearing Viewport<'a> that stays Copy. No wire schema or protocol change; the GPU path is Stage 3. 48 acceptance tests on the real paint_frame grid, every behavioral claim bite-verified. Durable design points, each a trap Stage 3 inherits:

  • the map's unit is a merged hidden component (overlapping or adjacent intervals unioned, keeping the earliest visible head), not a fold — folds may cross, and a later fold's own head can be hidden;
  • instances are per rendered window and per command/event operation, never per frame; a command's map follows the operation's target window, since a wheel event names a pane without activating it;
  • fold projection is per-frontend (FrontendView.fold_projection) — shared EditorCore motion would otherwise make a simultaneous unfolded GPU session's cursor skip lines it still displays;
  • a hidden cursor normalizes by position, not row, and set_view_top clamps in the setter rather than being repaired at render time;
  • the interactive-Lua unfold keys on the post-intercept edit site — a managed buffer intercept may legally relocate the op.

  Process notes worth keeping: main moved under the arc, and the merge was textually clean but not semantically clean (#146 added Viewport literals the new folds field invalidated) — a clean git merge-tree does not mean the merged tree compiles. CI was red at review on the macOS/luajit outline_5_level_100_entry_renders_within_100ms budget flake and went green on rerun.

• Documentation ledger refresh — MERGED as #147 (main @ 0a479ae, 2026-07-24). The #142 housekeeping, expanded after review found the ledger stale through four merges rather than one. Its own macOS/luajit red was the vterm VTERM_ALT_READY PTY timeout; green on rerun.

• Web grammars HTML + CSS — MERGED as #146 (main @ 47581f4, 2026-07-23). .html/.htm/.xhtml and .css highlight off the official tree-sitter-html 0.23 / tree-sitter-css 0.25 crate query constants (no in-repo overlay), and HTML's INJECTIONS_QUERY lights up <script> → js and <style> → css. Durable lesson recorded in docs/web-grammars-html-css-framing.md: the highlight.rs capture table is global, so adding a capture name retro-paints every other language — check the reverse direction and pin it.

• LaTeX Stage 1 — MERGED as #144, with its parent inline-math framing committed as #145 (main @ f09b0a1, 2026-07-23). .tex/.latex/.sty/.cls highlight via codebook-tree-sitter-latex 0.6 plus the first in-repo query overlay (builtin/queries/latex/highlights.scm, include_str!) — the reusable pattern for grammars whose crate ships no usable queries. The crates.io tree-sitter-latex is provably broken (no scanner.c). The math parser and Tiers 3–4 are deferred to the inline-math arc.

• Folding Stage 1 (headless fold engine) — MERGED as #142 (main @ c49a8c7, 2026-07-23, after three review rounds; round 3 clean). The instance-side fold store + translating/dropping View, the structural source (derived head line, closer-aware tail), the Lua data API + interactive C-c @ commands, the command-path pre-edit unfold, and authoritative-empty FoldState production landed with no protocol bump. The folding branch and worktree (../pmacs-folding) are retained but carry nothing unmerged; the folding-framing.md framing is preserved. CI red at merge was an unrelated environmental perf flake (outline_5_level_100_entry_renders_within_100ms, macOS/luajit only), green on rerun. Stage 2 has since merged as #149 (above); durable substrate seams live in docs/agent-handoff.md §1.

• Vterm Stage 3 (protocol v19 + GPU terminal) — MERGED as #135 (main @ cac4961, 2026-07-22, after two review rounds). Arc 5's terminal stage is complete (compile mode #113, Stage 1 #126, Stage 2 #130, Stage 3 #135). Its lane, worktree (../pmacs-vterm-gpu), and branch are done; durable substrate facts live in docs/agent-handoff.md and docs/vterm-framing.md.

• Branches deleted 2026-07-22 (authorized): vterm-stage3-framing (Revision 8 framing; its content is carried on vterm-gpu, verified as a superset before deletion — the branch was NOT an ancestor of vterm-gpu because the framing was copied rather than merged, so it needed a forced local delete) and tab-width-parity (a clean ancestor of main via #137). Both removed as worktree + local ref + githubsucks ref; the origin tracking refs were pruned. The -framing branches for each are deliberately kept.

• Tab-width rendering parity — MERGED as #137 (main @ 2625ec7, 2026-07-22). One fixed 8-column TAB_STOP_COLUMNS in pmacs-protocol now drives core/TUI columns, GPU code projection, and minimap width; source bytes and protocol ranges are unchanged. Its lane, worktree, and tab-width-parity branch (local + githubsucks) are deleted; the tab-width-parity-framing branch is kept. This closes the long-standing "tab width is a rendering-parity bug, NOT a config gap" deferral recorded in docs/agent-handoff.md §5.

• Locals-query processing — MERGED as #134 (with handoff #136), and modeline detection handoff #133. Both landed between this lane's base and its canonical-main integration.

• Config registry — MERGED as #127 (main @ 2e37c04). Its lane (config-registry, worktree ../pmacs-config-registry) is done; the branch is kept but carries nothing unmerged. Durable substrate facts moved to docs/agent-handoff.md §1 per rule 3 below.

• Both this and Vterm Stage 1 ran as concurrent lanes in sibling worktrees off main, with the shared files (src/editor.rs, src/lua_bindings/mod.rs, src/lib.rs) assigned to one lane each in advance. The rebase of the second lane onto the first had zero conflicts — worth repeating for future parallel work, along with its precondition: agree the file split before either lane starts, and keep each lane's footprint in the other's files to a single line.

Update protocol

Whenever a listed lane changes materially:

1. update its public branch and head/state here;
2. record new verification and remove superseded caveats;
3. keep durable architecture in docs/agent-handoff.md, not here;
4. remove the lane after merge or abandonment;
5. verify every recovery command from a clean worktree before calling the transfer complete.