An instrument, not a gate. Seeds a busy channel against the mock bridge and
measures the main-thread cost the headless correctness suite cannot feel.
Two measurements:
- Fast-wheel scroll of the bounded ~200-row window: compositor-cheap
(~0.2ms layout + 1.3ms recalc over a 241-frame, 12k-px burst).
- Prepend re-render cost while scrolled up: ~9-13ms main-thread tick,
attributable to the O(rendered-rows) parent walk (videoReviewContext
rebuild + day-group boundaries + element construction), NOT layout
and NOT leaf-row reconciliation (MessageRow memo bails correctly).
Anchor holds at 0px drift on this path.
Scope limit: measures Chromium reconciliation, not the WKWebView
compositor feel — that remains the real-Tauri macOS pass.
Run: pnpm build && npx playwright test --config=playwright.perf.config.ts
Co-authored-by: Tyler Longwell <tlongwell@squareup.com>
Signed-off-by: Tyler Longwell <tlongwell@squareup.com>
Scrolling up fetched a fixed 100-message batch and stopped. Because thread
replies collapse into their parent's summary and non-content events never
render, a reply-heavy window could fetch 100 events but grow the timeline by
only a handful of rows — making one wheel-up feel like it did nothing.
fetchOlder now pages in batches until at least MIN_TOP_LEVEL_ROWS_PER_FETCH
(10) *visible* top-level rows have been added, or history is exhausted. A new
pure helper, countTopLevelTimelineRows, counts rows the same way
buildMainTimelineEntries renders them: content kinds, minus deletions, top-level
or broadcast replies only. The dedup/exhaustion guards already in place
terminate the loop when the window stops advancing.
Co-authored-by: Tyler Longwell <tlongwell@squareup.com>
Signed-off-by: Tyler Longwell <tlongwell@squareup.com>
Each day group renders as a `<section>` whose React key was the exact
`createdAt` of its first message. Scrolling up prepends older messages,
and when a batch landed on a calendar day already on screen, the first
message of that day changed — flipping the section key. React can't diff
a changed key, so it unmounted and remounted the entire day section, all
its rows torn down and rebuilt above the reader's eye. That whole-section
remount is the residual flicker on scroll-up: the anchor restore was
correcting a full teardown instead of a clean prepend.
Key the section by the local start-of-day of its messages, which is
stable across same-day prepends, so an older message grows an existing
section's children (stably-keyed rows reorder, not remount) instead of
replacing the section. Fold the duplicate key derivation in the render
loop into the lib boundary's `key`, which was already documented as
"stable" but wasn't.
Pure helper `startOfLocalDaySeconds` plus lib tests for day-key stability
across a prepend and separation across calendar days. tsc, biome, 40 lib
unit, scroll-history e2e 6/6 green.
Co-authored-by: Tyler Longwell <tlongwell@squareup.com>
Signed-off-by: Tyler Longwell <tlongwell@squareup.com>
The older-history fetch spinner renders above the anchor row but toggles on its own render commit (messages unchanged). The anchor-restoration layout effect was keyed only on `messages`, so it never re-ran when the spinner appeared or disappeared — leaving the spinner's height as an uncorrected shift above the reader's eye, the residual flicker on prepend. Thread isFetchingOlder into useAnchoredScroll as an extra restoration trigger so the existing scrollBy correction fires on the spinner toggle too, making the anchor the single owner of every layout change above the fold.
Co-authored-by: Tyler Longwell <tlongwell@squareup.com>
Signed-off-by: Tyler Longwell <tlongwell@squareup.com>
ChannelRouteScreen splices a getEventById-fetched target into targetMessageEvents so a deep link works in a channel whose feed doesn't already contain the message. The fetch effect wiped those events whenever the route target cleared — and onTargetReached clears the messageId URL param the moment the row is centered — so the only copy of the deep-linked message vanished and the timeline went blank.
Reset spliced events on channel/forum-post change only (guarded effect keyed on channelId/selectedPostId, seeded with the mount key so first-commit cache seeds survive), and have the fetch effect merge cached/fetched events additively instead of overwriting. e2eBridge: route the mock-engineering-shipped known event by 'h' tag so getChannelIdFromTags resolves it.
Co-authored-by: Tyler Longwell <tlongwell@squareup.com>
Signed-off-by: Tyler Longwell <tlongwell@squareup.com>
useAnchoredScroll is meant to be the single owner of scrollTop, but both
live scroll containers had lost the [overflow-anchor:none] class (it
survived only on the thread loading skeleton, which never scrolls). With
it gone, Chromium's native scroll-anchoring heuristic re-engaged and
applied its own scrollTop correction on prepend — picking its own anchor
element — while the hook applied a second scrollBy correction on top. When
the two anchors diverged the corrections stacked, producing the residual
jiggle that survived the single-owner rewrite.
Restoring the class on both live containers (MessageTimeline timeline +
MessageThreadPanel body) hands scroll ownership back to the hook alone.
This is a real-wheel-only symptom: native scroll anchoring barely fires
under synthetic scrollTop= writes, so the headless suite stayed green
while a manual macOS scroll pass surfaced it. scroll-history e2e 6/6,
tsc clean.
Co-authored-by: Tyler Longwell <tlongwell@squareup.com>
Signed-off-by: Tyler Longwell <tlongwell@squareup.com>
Replaces the two competing scroll writers (useTimelineScrollManager +
useLoadOlderOnScroll) with one anchor-based primitive, useAnchoredScroll,
owned by both the main timeline and the thread panel.
The prior design had two hooks both mutating scrollTop on prepend, which
is the root of the timeline-jumping bug: a fetch-older restore and a
scroll-anchor adjustment would fight over the same frame. The new hook
keeps a single anchor (the row the reader's eye is on, picked by a
top-crossing walk) and restores it relative to its prior top offset after
every render — prepends, appends, image loads, and embed expansions all
flow through that one path.
Also fixes three concrete bugs surfaced by the ported E2E suite:
- Deep-link targets in older history: the post-mount target effect bailed
once when the row wasn't loaded yet and never retried. It now keys on
`messages` and re-runs until the spliced-in row commits, then centers.
- Root-message deep-link reopen: the initial-mount target path scrolled
and marked the target handled but never fired `onTargetReached`, so the
`messageId` URL param stuck and re-clicking the same link was a no-op.
The initial path now fires the callback too, matching the post-mount one.
- Inline image height reservation: images never read their NIP-92 `dim`
tag, so a tall image grew from ~0 on load and shoved the timeline. We
now stamp intrinsic width/height so the row reserves space before decode.
Verification: tsc clean, biome (764 files) clean, 989/989 unit tests,
scroll-history 6/6, full e2e smoke 253 passed (3 failures all reproduce
on a clean main checkout — pre-existing, unrelated to this change).
Co-authored-by: Tyler Longwell <tlongwell@squareup.com>
Signed-off-by: Tyler Longwell <tlongwell@squareup.com>
