* fix: stop iOS focus-zoom and casual pinch-zoom on mobile
* fix: scope mobile 16px rule to editable controls, document zoom tradeoff
* fix: keep code-block language picker compact on touch devices
* test: pin viewport assertion to exact maximum-scale token
* fix: keep code-block language select under mobile 16px zoom floor
* fix: apply mobile 16px zoom floor on hybrid touch devices via any-pointer
* fix: drop maximum-scale=1 viewport lock (no-op on iOS, fails a11y audit)
* feat(web): add sidebar view system to mobile task switcher
* test(web): scope mobile view-system gear indicator to the sheet
* fix(web): disable view-chip drag-reorder on touch to keep row scrollable
* fix(web): use mouse+touch sensors so chip drag works on hybrid devices
* fix(web): match kanban PointerSensor+TouchSensor convention for chip dnd
* fix(web): use MouseSensor+TouchSensor so touch swipe scrolls chip row
* fix(web): wire group collapse + align/clean view bar in mobile sheet
* feat(web): surface PR mergeability + conflict state in the CI popover
The CI hover popover showed checks/reviews but nothing about mergeability, so an
unmergeable PR silently dropped its merge button with no reason. Add a mergeability
row that reuses the detail panel's conflict banner + "Resolve conflicts" agent prompt
for dirty PRs, a "Behind base" chip, and a branch-protection note explaining why a
blocked PR is gated. Replace the compact merge split-button+dropdown with an inline
primary + "or <method>" buttons so the menu's detached portal no longer closes the
hover popover.
* feat(web): reflect PR mergeability in status colors and badges
The PR status pill, badge, and chat-input CI chip derived their colour from
checks/reviews only, so a conflicted or blocked PR still read as a green "all good"
check. Map dirty -> red, behind -> amber, and branch-protection blocked -> amber across
getPRStatusColor, the PRStatusIcon badge, and the chip glyph, so the indicators agree
with the popover. Blocked stays sky/in-progress while a requested review is still
outstanding (that's the awaiting-review gate, not a hard block).
* refactor(web): remove dead merge-menu keep-open plumbing
The compact merge control was redesigned to inline buttons (no dropdown portal),
which already fixes the hover popover closing. That leaves the keepOpen ref /
setKeepOpen / onMergeMenuOpenChange / onMenuOpenChange chain threaded through
usePopoverInteractions, PRCIPopover, and PRMergeButton unreachable: the popover
always renders PRMergeButton with compact, and the only DropdownMenu lives in the
non-compact detail-panel path that never passes the callback. Drop the unused
scaffolding (flagged by OpenCode + greptile).
* fix(web): keep CI popover quiet for blocked PRs only awaiting review
PRMergeabilityRow rendered "Blocked by branch protection" for any blocked PR,
including one blocked solely because a requested reviewer hasn't responded. That
contradicted the chip/pill/badge, which deliberately read that case as the calm
"awaiting review" state. Skip the note (render nothing) when isPRAwaitingReview is
true, and narrow blockedReason so an empty checks_state no longer falsely claims a
required status check failed. Flagged by Claude (blocker) and greptile.
* refactor(web): rename mergeability-notice compact prop to popover
The `compact` prop made the chip/text larger (to match the popover's text-xs rows)
while ConflictBanner's `compact` tightened padding — so "compact" rendered bigger,
which reads as inverted. Rename it to `popover` (the variant rendered inside the CI
hover popover) and document that the chip intentionally matches the popover rows.
No visual change. Flagged by cubic.
* fix(web): show default cursor on disabled compact merge alternates
The compact "or <method>" alternates kept cursor-pointer while disabled (during the
~1s merging state); add disabled:cursor-default to match the primary button. Flagged
by OpenCode.
* fix(web): suppress kanban spinner for TODO tasks
A task moved back into TODO can still carry an active primary session
state (CREATED/STARTING/RUNNING). The kanban card was painting the
blue running-spinner in that case, making queued tasks look like they
were executing.
Add an explicit TODO guard at the top of shouldShowTaskRunningSpinner
so the queued column never shows the spinner, regardless of the
session-state desync. Existing decoupling for IN_PROGRESS/REVIEW/etc.
is untouched.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* test: clarify spinner test description excludes TODO
Greptile flagged that "returns true when any task state has an actively
running primary session" overstates the invariant now that TODO
short-circuits to false. Reword to "non-TODO task states" so the
description matches behavior.
* ci: retrigger after artifact expiry
---------
Co-authored-by: Kandev Agent <agent@kandev.dev>
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
* perf(backend): cap cumulative diff bytes and file count during large rebases
* perf(web): coalesce cumulative-diff refetches and skip redundant status compares
* feat(web): warn when the changes panel hides files from a capped diff
* fix(backend): enforce cumulative diff total byte budget strictly
* fix(web): keep coalesced refetch timer shared across envKey subscribers
* fix(web): drop unsound git-status SHA fast-path that staled live diffs
* fix(web): cancel coalesced timer on mid-flight drain to avoid duplicate refetch
* refactor(web): move changes-panel banner and helper to a sibling file
* docs(backend): clarify cumulative diff truncated-files-count semantics
* fix(web): tighten useSelectedFileKey mode param to the all|file union
* fix(web): stop session-tab flicker when same-task sessions have diverged env ids
* test(web): fail fast on missing dockview bridge and clear diverged-env interval
* perf(kanban): memoize derived board data to stop re-render storms
The board rebuilt the `tasks` array (and `stepOptions`/`sharedProps`) on every
store change, producing fresh object identities that defeated downstream memos
and re-rendered every card on each WebSocket message. Wrap them in useMemo so
the board only re-renders when its inputs actually change.
* perf(session-runtime): bound runtime memory and dedupe git-status compare
Three runtime-store fixes for the crash a heavy rebase triggered:
- Cap shell + terminal output buffers at 2MB (matching the existing process
buffer) so a chatty shell can't grow the store without limit.
- Purge per-session runtime state (messages, turns, shell/process buffers, and
the bySessionId maps) when a session is removed; env-scoped maps are dropped
only once no other session shares the environment.
- Run setGitStatus's deep per-file diff comparison once per update and return
whether anything changed, instead of recomputing it 2-3x. Under a massive
rebase (thousands of files, frequent updates) that comparison was the
dominant CPU cost and drove the render storm.
* refactor(session-runtime): drop now-unused gitStatusWouldMutate
The git-status handler no longer pre-checks for changes (setGitStatus now
returns whether it changed), leaving gitStatusWouldMutate with no callers.
* fix(session-runtime): enforce terminal cap on first and oversized chunks
appendTerminalChunk only evicted when more than one chunk was buffered, so a
single chunk larger than the 2MB cap slipped through untrimmed; the new-terminal
branch also pushed the first chunk without going through the cap at all. Clamp
each chunk to the tail of the cap and route the new-terminal path through
appendTerminalChunk. Addresses CodeRabbit review.
* fix(web): restore diff expansion for committed-source review rows
#1097 narrowed the expansion gate to `source === "uncommitted"` to dodge a
trailing-context crash, which silently dropped expand controls for committed
files — the bulk of a review. The crash is now caught by the reparse guard in
useExpandableDiff, so re-enable expansion for committed rows, fetching the old
side from the repo's base branch (not HEAD, which already has the commits).
Extracts the decision into a pure `resolveDiffExpansion` helper and adds a
multi-repo guard so a committed file never borrows another repo's base branch.
* fix(web): guard diff-expansion base fallback to single-repo tasks
Resolve baseBranchByRepo once in ReviewDiffList (not per row) and only expose a
fallback base branch when the task has exactly one repo, so a committed file
with no repository_name in a multi-repo task no longer borrows an arbitrary
repo's base branch. Addresses Greptile/cubic review feedback.
Carlos Florêncio