Review follow-up: the llm/adapters-updated listener's disposer was
discarded, leaving it firing (harmlessly, behind isDisposed()) between
TUI shutdown and fiber disposal, asymmetric with the sibling channel
listeners. The controller now exposes detach(), and the channel's
detachListeners() calls it on both the dispose() and startup-failure
paths.
The card is a React child of the wrapper, so capture-phase presses on it
reached the wrapper's dismissal handler: the first pointerdown of a text
selection closed the card, contradicting its JSDoc contract. Restrict the
immediate close to presses outside the card, keeping it mounted under a
held press (and the browser click with it), and align onPointerLeave's
grace arming with Menu (only while open).
Pin both new behaviors in hover-card.spec and update the bilingual Agent
Note.
Review flagged that "equal by construction" rested on an engine behaviour I
had not measured: `scrollbar-gutter: stable` only equalizes the layers if the
engine applies it to `overflow: hidden` the way it does to `overflow-y: auto`.
Measured it on the running app across the three engines Playwright ships,
and the property does not hold up.
engine .input / .backdrop / .mirror wrap width
chromium 776 / 776 / 776 (768 / 768 / 768 with the declaration)
firefox 776 / 776 / 776 (unchanged by it — overlay scrollbar)
WebKit 768 / 776 / 776 (unchanged by it)
WebKit reserves for `overflow-y: auto` and not for `overflow: hidden`, so the
declaration left .input at 768 against 776 — exactly the gap it was meant to
close — on the one engine where that gap is observable at all, while costing
every chromium user 8px of text column unconditionally. Reverted: the
composer's metrics are now the same as before this PR.
The WebKit gap predates this change and is not closed here. It is recorded in
the Agent Note with the numbers, and the browser scenario asserts the equality
on the lane's engine so a regression into that state fails loudly. The mirror
is unaffected on WebKit for the drafts measured — the extents still agree — but
a draft whose wrapping turns on those 8px would clamp it.
The review's monotonicity concern resolves the same way: the declaration was
never worse than master, because WebKit already measured 768 against 776
without it. It simply was not better.
Also from this round: the wrap-width assertion now covers .mirror as well as
the two glyph layers — it is the height authority, so a mirror alone wrapping
wider would measure the box short and clip content below the 14-line cap with
every other assertion green. Plus `renderGeometry`'s missing `@param
trailingNewline`, and both e2e tsconfig lists restored to alphabetical order.
Second review round escalated the wrap-width divergence from a separate
concern to a defect in this fix's own premise, and it is right.
Only .input scrolls, so only .input loses content width to a scrollbar that
consumes layout space — what Windows and Firefox draw, and what the theme's
global `::-webkit-scrollbar` width makes chromium treat as occupying space.
A narrower .input wraps a long soft-wrapped draft onto more lines, so it
grows taller, its scroll maximum exceeds the backdrop's, and the mirrored
offset clamps below the caret. That is the same failure the trailing-line
sentinel fixes, in the same direction, so deferring it would have shipped a
fix that does not hold where users run a classic scrollbar.
My first attempt to reproduce it found nothing and was wrong: the probe
content was not wrap-sensitive. With varied-length words the effect is
plain — the same draft laid out at 8px-apart widths differs by 2 to 5
lines, while at equal widths a textarea and a div agree exactly.
The three layers now reserve the gutter together, in the shared metrics
block that already exists to keep them symmetric. `overflow: hidden` is
still a scroll container, so the non-scrolling layers honour it: 8px is
reserved on each, measured. The cost is a text column 8px narrower on every
platform, which is the price of one geometry rather than a per-platform one.
The browser scenario asserts the premise directly — equal wrap widths, and a
reserved band greater than zero on each layer. The band is what stops the
assertion being vacuous: the widths would also match with no reservation at
all on this engine's overlay scrollbar, and it is the reservation, not the
match, that carries the guarantee to a platform whose scrollbar takes real
width. Removing the declaration fails it with `expected 0 to be greater
than 0`, and fails the golden with it.
Also from the same round, three comment corrections: the e2e file header no
longer describes the deleted layout effect, the measurement guard no longer
claims the backdrop holds exactly one text node (the sentinel makes a
second), and the sentinel comment now carries the one-sidedness argument
that also settles the ghost hint — the mirror only fails when the backdrop
is SHORTER, and the hint can only add content, never remove a line box.
Review caught a real divergence the earlier measurements missed: mirroring
an offset is only correct while both layers can reach it, and for a draft
ending in a newline the backdrop could not.
A textarea reserves a line box for the caret after a final newline.
`white-space: pre-wrap` collapses a text node's trailing newline and
generates none. So a draft ending in a newline made the backdrop exactly one
line shorter than the textarea — measured 628 against 652 — and the mirrored
assignment clamped, leaving the glyphs one line behind the caret at the very
bottom of the draft.
The backdrop now carries the same trailing-line sentinel the mirror div has
carried all along: its content is the decoration walk plus one newline. The
same pre-wrap collapse absorbs it when the draft does not end in a newline,
so it costs no height in the ordinary case, and it supplies the missing line
box when it does. Verified in isolation first: a bare pre-wrap div measures
180/180/198 against a textarea's 180/198/216 for zero, one and two trailing
newlines, and 180/198/216 with the sentinel.
Coverage for the shape that exposed it: the browser scenario asserts the two
extents are equal before asserting the glyphs reach the end, observing each
layer's maximum by asking for an impossible offset and reading back the
clamp rather than computing it from scrollHeight, and the golden records the
relation. The unit spec pins the backdrop's text as the draft plus exactly
one newline. Removing the sentinel fails both, the e2e with the same 628
against 652.
The scrollbar-gutter half of the same review point does not reproduce here:
both layers measure clientWidth 776 against a border box of 776 while the
draft overflows, so this engine's textarea scrollbar is an overlay and takes
no width out of the wrap.
The first version coupled the layers from two places: a `scroll` listener
and a layout effect keyed on the committed draft. Mutation-testing each
hook alone against the built client shows the effect never fires the only
assignment that matters — with just the layout effect disabled the browser
scenario stays green, while disabling just the listener fails it.
Both premises behind the effect were wrong. Typing scrolls the caret into
view, which is an ordinary `scroll`. A draft that shrinks past the current
offset clamps both layers to the same maximum, because their extents are
equal — measured in chromium at 964/964, 1012/1012, 844/844 and 820/820 for
plain, soft-wrapped, unbreakable-run and highlighted drafts — and the
textarea's clamp fires `scroll` too.
The hazard the effect was imagined to cover does not exist either: React
replacing every child of the backdrop when the decoration set changes shape
preserves `scrollTop` (measured: 300 stays 300 through a full child
replacement), and the only replacement that zeroes it shrinks the content
below the offset, which is the clamp case already covered.
The e2e's edit case survives, retitled to say what it actually pins: that
typing is not a separate case needing its own mirror. The unit spec now
asserts the backdrop tracks a second move back to the top, which a one-shot
mirror would fail.
A composer draft past the 14-line cap could not be scrolled: the caret and
the selection moved, but the words stayed frozen at line 1, so the tail of
anything longer than the cap was unreachable while writing it.
The composer paints its text in two stacked layers. The textarea owns the
value, the selection and the caret but renders its own glyphs transparent;
every visible character is painted by the decoration backdrop beneath it,
which also carries the claim-token highlight, the chips and the ghost hint.
The backdrop is `inset: 0; overflow: hidden` — clipped, not scrolled — and
nothing linked its offset to the textarea's. Below the cap both layers rest
at 0, which is why the defect hid behind every short-draft screenshot and
fixture.
InputBar now mirrors the textarea's scrollTop onto the backdrop, from a
`scroll` listener (every gesture and every caret-driven scroll) and from a
layout effect keyed on the committed draft (an edit reflows both layers
without necessarily firing a scroll event).
Scrolling is layout, so jsdom cannot show this: the unit spec stubs both
offsets and proves the mirroring paths run, while a new browser scenario
measures the user-visible fact against the built client with a DOM Range
over the backdrop's own text — after a wheel gesture over a 40-line draft
the last line is on screen and the first has scrolled out. Confirmed both
directions: with the mirroring reverted and the packages rebuilt, the
golden reads `last draft line is on screen: false` while `textarea moved:
true`.
Loader activation is service-driven, so the TUI can mount before a
configured adapter plugin registers its provider routes; every fresh
session then printed 'Could not resolve model context: no adapter
registered for provider …' for a working configuration.
The model controller now treats a NO_ADAPTER rejection of the
context-window resolution as transient: it parks the resolution
silently and re-resolves on the next llm/adapters-updated commit. A
commit that still lacks the route parks the wait again; any target
change clears it; all other resolution errors still surface. A wrong
provider name keeps failing loudly at dispatch, where it is actionable.
The workspace browser's two hover-raised popups both died on the way to
them. HoverCard closed on the first pointerleave and rendered its card
pointer-events:none, but the card sits 8px off the anchor, so every path
to it crossed ground belonging to neither. The row action menus put
closeOnPointerLeave's handler on the portaled list, so aiming back at the
... trigger that opened it, or overshooting a list edge, closed it with no
window to come back.
usePointerGrace owns one cancelable delayed close (200ms) shared by both
atoms: leaving arms it, returning cancels it. The hover card becomes
hit-testable so resting on it holds it open, and Menu moves pointer-leave
dismissal to the wrapper span, where React's enter/leave traversal makes
trigger and portaled list one region.
Both gestures are pinned in the real browser lane; each fails without the
corresponding fix.
Review follow-ups on the approval takeover:
The scroll region is now a tab stop (tabIndex + named role="group"). The
question composer's scroll body needs none — its option rows are focusable and
pull the container along — but this one holds nothing but text, so a
keyboard-only user could reach the buttons and never the command's tail, and
approve what they could not finish reading.
The action row's padding reproduces the 14px gap it had inside the body: the
flex gap of 6 plus its 8px top margin, neither of which reaches it now that the
row sits outside the scroll region. The resting card is unchanged again.
The answered-transcript golden captured the OS's own refusal of the denied
first attempt — "bash: notes.txt: Operation not permitted" on macOS against
"bash: line 1: notes.txt: Read-only file system" on Linux — so it passed
locally and failed the Linux snapshot lane.
The scenario now keeps one golden (the waiting panel, platform-neutral) and
asserts the answered state on the world instead: the decided outcome, the file
the escalated command actually wrote, DONE, the panel gone, and the composer
re-enabled. The file assertion is stronger evidence than the transcript dump
it replaces — it proves the grant reached the executor.
- .loadingFloat moved after the .status/.error block (its padding was
losing the same-specificity race) and re-anchored bottom-right: the
truncated/error rows own the bottom left and keep rendering through a
scan, so the pill can never cover them; confirmCreate's relist now
clears the stale failure text like every other scan launch.
- The close edge resets loading, so the slow-scan effect disarms while
hidden and a reopened dialog waits out a fresh silence window
(regression test added).
- The truncated note's survival through a scan is now asserted in the
slow-scan test; the wait-bound test moved to fake timers with the
200ms bound explicit.
- select()'s exemption from the one-frame rule and the constants' local
calibration premise are recorded in JSDoc and the capability-seam
Agent Note; the themed-scrollbars note's rebinding enumeration is
replaced by a pointer to the mechanical gate (it had drifted twice).
Both pairs re-recorded.
The approval panel replaces the InputBar while a sandbox escalation waits,
and its justification and command are unbounded model text. With no height
cap, a long command grew the card until the refuse/allow row went under the
fold: at 900x700 the action row's bottom landed at y=749, so the user could
read the request and not answer it.
Justification and command now scroll in one region capped at the same height
as the composer's draft area, with the amber strip and the action row outside
it. The cap is one value with two consumers — declared as
--dsh-composer-text-max-height on ConversationRoot's .composerSeat, the
composer chain's only shared ancestor — so the seat cannot cap its two states
differently. The card rebinds the l2 scrollbar pair like every other scrolling
surface on an elevated background.
Covered by a new web e2e scenario that drives the real composition (read-only
session, denied write, the model's escalation retry, answer clicked through
the panel) and measures the live panel at two viewport heights against the
composer's own cap, read off the textarea rather than hardcoded.
Move COMPACT_CHECKPOINT_SOURCE and isCompactCheckpointSource into a
cordis-free src/checkpoint.ts leaf, re-exported from the root so every
host-side consumer keeps its import. The client can then type-import the
leaf without reaching dsh-session's root, whose Context merge declares the
host sessions service and collides with the client's -- the dsh-commands/brand
shape. Renaming the plugin id now fails the client typecheck.
Also: keep recoverable summary text when a compact/summary mixes text with
other block types, and capture the seeded-history provenance seqs from the
pushes that produce them instead of deriving them by arithmetic.
Replace the surface-ordered fold with a log-ordered human transcript:
append-origin surface events at their own log positions plus one marker per
landed compaction checkpoint. Command folding, the tool-call index, and the
rev-keyed memo carry over unchanged.
Removes foldDegraded, the padding sentinels, baseSeq, and degradedSeqs() --
they existed only to satisfy the core fold's seq === index assertion. That
also closes the pagination hole A1 exposed: a page can carry a checkpoint
whose shadowed range fell outside the window, and nothing resolves
surfaceOp.start anymore.
Conflicts, all in files this branch and master both touched:
- `chat/helpers.ts` — kept both sides. Master replaced the inline env
filter in `gitBranch` with `scrubbedParentEnv()`; this branch added the
surface-marker and compaction-source imports.
- `docs/module-graph.md` — regenerated. Both sides added a `tui` edge
(master `subprocess`, this branch `compact`); the generated row now
carries both.
- Three `.i18n.yaml` pairing records — re-recorded from the merged files.
Blob hashes cannot be hand-merged, and both sides' prose survives on
each side of every pair.
`docs/cordis-catalog/services.md` also regenerated for shifted source
lines and master's new `sessionTitle.rename` entry.
Marker scale was deferred in a resolved review thread but only the
compaction-progress half was written down. Deferred now names it, says
where the count comes from (`sourceEventSeqs`), and why it belongs with
progress rather than here.
Adds the prerequisite: the terminal's replay and live replacement branches
are textually identical and 600 lines apart, so marker content needs one
home — fold them into a single `renderReplacement(event)` before giving the
row a payload that must stay consistent across both paths.
Review follow-ups on b9b2e593f, all documentation precision.
The redundancy note on `isCompactCheckpoint` leaned on a reading its call
site does not state: `index.ts` reaches it for surface-eligible non-append
events, which is the same set as replacements only because the marker is
mandatory. Say that instead.
The Agent Note now owns three facts it was leaving to a future reader.
`rebuildTranscript` materializes a component per append-origin event and
runs on mount, color-scheme change, and every reasoning toggle — work
compaction used to bound for exactly the long sessions it serves, so the
cost now tracks session length rather than the surface. `Consequences`
names `surface-replayed-compaction` as the durable evidence for the
live/replay equivalence claim, so the two fixtures that must move together
are findable from the Note rather than the PR thread. `Deferred` records
that a page can now carry a checkpoint whose `surfaceOp.start` fell out of
the window: pagination no longer cuts on the checkpoint's provenance
group, `FoldAdapter` pads with a non-surface sentinel, and `nodes()`
degrades to `degradedSeqs()` — which is already close to the transcript
projection A2 should build deliberately.
Also corrects the definite-assignment comment in the snapshot scenario:
the assertion rests on the awaited setup invoking `beforeMount`, not on
that call being synchronous.
Review follow-ups on the append-origin transcript projection.
The live/replay equivalence claim was stated unconditionally but does not
cover `tool/call`: only replay re-derives call pairing, because a call
event carries no `surfaceOp` of its own and inherits transcript
membership from the `assistant/message` that advertised it — which the
live listener has necessarily just rendered. Narrow the claim in the TUI
README and Agent Note, and record at `rebuildTranscript` why the filter
is replay-only rather than a missing live branch.
Add `surface-replayed-compaction`: the three existing fixtures all come
from the live path, leaving the resume case the bug report leads with
pinned only by a unit test. The new checkpoint mounts with the
replacement already stored and records byte-identical to
`surface-after-compaction-wide`, so the two fixtures now pin the
equivalence they assert. The shared fixture appends move into
`appendPreCompactionLog` / `appendCompactionCheckpoint`.
`MESSAGE_TYPES` is not "human message event types" — it includes
`assistant/message`. Say what the code distinguishes (append-origin
conversation messages vs. model-only replacement copies) at the const,
the `paginate` and `session.history` JSDoc, the apiproxy README, and the
Agent Note.
Also: spell the replace shape as `Extract<SurfaceOp, { op: 'replace' }>`
for symmetry with the module's two other uses; document why
`isCompactCheckpoint` keeps a replacement check that is redundant at both
call sites; say that Ctrl+R toggles reasoning, which rebuilds the
transcript; and qualify "the sole source of derived history" as derived
*model* history now that the transcript is the other projection.
The terminal and history pagination both treated the model-visible surface as
the human transcript. A landed compaction replacement therefore erased the
conversation it summarized — messages the reader had already seen — and a
model-only replacement copy consumed a page's `maxMessages` quota, which could
also split a compaction's provenance from the replacement citing it.
`dsh-session` now exports the marker split `isAppendSurfaceEvent` /
`isReplacementSurfaceEvent`. The terminal replays append-origin surface events,
keeps a shadowed step's tool cards paired through its append-origin assistant
message, and renders one dim marker where a compaction landed; the checkpoint is
recognized through the compaction seam's `isCompactCheckpointSource` contract,
not the shape of the replacement. `session.history` counts only append-origin
human messages. Everything model-facing keeps reading `session.surface`.
creatixchu