Waking card, owned and paired surfaces, the asleep view and the Sleep-failed toast spliced "This
workspace" into {{name}} templates ("Couldn’t wake This workspace", "Waking This workspace.").
cloud-workspace-name-copy.ts gives each sentence an unnamed variant ("Couldn’t wake this
workspace", "Couldn’t put this workspace to sleep", "Waking this workspace."…) and headings read
"Cloud workspace". Drops the two private fallback keys it replaced.
* fix(cloud): SLEEP-IN-PROGRESS (2/2) a paired cloud workspace shows the waking overlay, its sealed read-only copy with Wake to continue, and the saving state after Sleep
- PairedCloudWorkspaceSurface, mounted from OrcaCloudWorkspaceSurface for paired rows
- the same CloudWakingOverlay (Retry wakes through the control plane, billing copy included)
- asleep: the control plane's sealed snapshot via useCloudReadSnapshot + CloudReadSnapshotView
- Sleep marks a paired copy saving; re-read every 2.5 s until a new-enough copy, failed at the deadline
- useWakeOverlay moved to its own module (shared by owned and paired)
* fix(cloud): SLEEP-IN-PROGRESS (2/2) re-read the sealed copy through the window-visibility interval (recurring-timer ratchet)
Operator routes GET /v1/internal/secrets/kek and POST /v1/internal/secrets/rewrap
move every envelope-encrypted row (secrets, agent_credentials, github_user_tokens)
from a retired KEK to the active one: idempotent, resumable, conditional writes,
counts-only logs. A v2 row format behind SECRETS_SEAL_FORMAT=v2 (default v1)
keeps the KEK id out of the value layer's AAD so rotation re-wraps data keys
without decrypting values. Adds the KMS design doc.
* fix(cloud): SLEEP-IN-PROGRESS (1/2) a sleeping cloud workspace says it is saving its read-only copy until the copy is in, and says so if it never arrives
From the Sleep press until a copy at least as new as the press is read back, the row ("Going to
sleep · Saving copy…"), the asleep view and the Files pane show a saving state instead of their
"nothing was saved" lines or an older copy; the view polls every 2.5s meanwhile. No copy within 90s:
"Couldn't save a read-only copy. Your files are safe in the workspace." with Wake. Owned workspaces
(paired ones have no read-only view yet: 2/2). Copy and layout per design-squad.
* chore(cloud): SLEEP-IN-PROGRESS drop the two now-unused localization keys
Restart reaches a paired workspace's sandbox: main's restart resolves a bound pairing
(restartPairedOrNotCloud) and restarts it through the control plane, dropping the bound server's
dead connection so it reconnects, with the same waking > connecting > ready / failed progress as an
owned restart. The panel offers Wake (the sidebar open path, which wakes paired rows since #571/#583)
and Restart on paired rows again; Archive and Move stay hidden until PAIRED-ARCHIVE.
packaged-cli-assets' macOS launcher test compares an os.tmpdir()-based path
with the realpath the launcher reports, so it failed in the gate on any PR
selecting it (#574 batch 8, parallel and serial). Export a realpath'd TMPDIR
so the gate environment matches what child processes see.
design-squad: arrow keys land on Retry like on any row, Enter runs it, and while it reads it shows
Loading… and can't run twice. Coordinator: the GitHub list (and other team answers) re-read on an
owner/member role switch; tests cover owner → member → owner, then a restart.
A failed first page resolved as an empty listing, cached for 5 minutes per identity, and an empty
list hides the Project picker's GitHub section: after a restart or role change the section was
gone for minutes. A failed read now isn't cached and the section says 'Couldn't load GitHub
repositories. Retry'; an empty answer isn't kept either, so each open asks again.
Every button that opens a URL in the external browser gets a chevron with
"Copy link" that copies the exact URL (minted the same way for lazy links),
for incognito windows or another browser profile.
* fix(cloud): PAIRED-WAKE opening a paired workspace right after its Sleep wakes it (fresh status read; paired Sleep refreshes the row)
n38 GATE: the first open after Sleep sat in "Reconnecting… retrying automatically" with a green
Active dot. wakePairedCloudWorkspaceIfAsleep read status through the 30s cache, and paired Sleep
never refreshed it, so the cache still said "ready" and no wake ran (a later open, past the cache,
woke it). The open now always reads status fresh; paired Sleep re-reads the row's status at once
(Asleep, not Active) and notes it slept here, so a failed read on the next open still wakes it.
* test(cloud): PAIRED-WAKE literal result types in the first-open mocks
* fix(cloud): PAIRED-CHROME paired rows get the live cloud mark, state word and hovercard (actions main can't run for them yet hidden)
* test(cloud): PAIRED-CHROME paired-row hovercard fixtures (running, asleep, menu)
* fix(cloud): PAIRED-CHROME asleep paired card says opening it wakes it (no Wake button until PAIRED-WAKE 2/2); split the card's primary action
#574 (deletes widely imported card files) selects ~52 batches; batch 1 failed on
three Electron browser-extension tests that pass alone and in full-suite runs.
A failed batch is now re-run once with --maxWorkers=1 and passes only if fully
clean (set -e), logged as 'failed in parallel, passed serially'. A real failure
still fails the gate.
Holding Option showed the delete quick action on local worktrees but never on a cloud
workspace's row. The gate (canShowWorkspaceDeleteQuickAction) refused every main
worktree, and a cloud workspace's own row is its sandbox's main checkout. The gate now
takes cloudPrimary (useIsCloudPrimaryRow: owned, or paired via the runtime catalog
or bindings). The live card's DeleteQuickAction (place.deleteQuickAction) reads it, and
the click goes through runWorktreeDelete, whose cloud-primary branch opens the cloud
delete confirmation (requestCloudPrimaryDelete), never Remove Project. A local repo's
primary checkout still gets no trash.
The lockfile still listed apps/eval (renamed apps/acceptance) and pinned jose 6.2.13 in three importers
while only jose@6.2.16 was in packages (a bad merge), so pnpm install --frozen-lockfile failed. Ran
pnpm install --fix-lockfile --lockfile-only: importer apps/eval -> apps/acceptance, jose -> 6.2.16
(the one already locked; satisfies ^6.2.13), two orphaned packages dropped. No other version changes.
Dev builds only: globalThis.__orcaCloudTokenClaims() in the main process returns { org, orgRole,
sub } from the current sandbox token's payload, never the token. Packaged builds install nothing.
The generator now emits VM_SCRIPTS_HASH and a monotonic VM_SCRIPTS_STAMP
(kept while the scripts are unchanged; max(previous + 1, now) on a change).
Provisioning records the stamp in /opt/orca/bin/.scripts-stamp (source
builds use the fetched commit's stamp), installEntryPoints writes the
Worker's, and ensureEntryPoints probes stamp + marker in one command and
skips images whose stamp is newer than the Worker's. Golden refreshes use
ensureEntryPoints too, so a lagging Worker cannot bake older scripts into
a refreshed golden.
A PR that deletes or changes widely imported modules selects hundreds of
test files; passing them all to one vitest spawn failed with E2BIG. Run them
in batches of PREMERGE_TEST_BATCH (default 150); every batch must pass, so
the gate and the premerge OK record are unchanged.