Repository navigation
fix(apps): hold Update under the local-only notice and link Update All to the trust step - #35990
Conversation
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS No design-level concerns. Each surface now offers the step that works, the only removal (the plain Update, hidden while the local-only notice is up) is listed with its reader, and five downloaded images show the reshaped surfaces. Notes
[DESIGN-REVIEWED] 86f7cda Description read[DESCRIPTION-READ] 8de53ae8f5f7fc17e373a6091f527ff9ca84af0adc2c008dbca329995300e0af That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
First Principles Review (Opus 5.5) — ✅ PASSPremise-level review of First-Principles-Verdict: PASS Nothing blocks: each item answers a rough spot that #35913 names or a measurement taken at 390x844, and the notice buttons use the existing What this change shipsInventory (6 items) — 6 justifiedIntent: once an update hits the local-only trust refusal, every App Store surface sends the user to the trust step that works. This is a FIX (#35913, follows #30979).
[FIRST-PRINCIPLES-REVIEWED] 86f7cda Description read[DESCRIPTION-READ] 408df86517dc445e4b2cfd3cd1762b6771c5233b0020bfd4fdd026dfd5143c26 That sha256 names the PR description this verdict read. Recompute it with |
Opus 5.5 Review — ✅ no blocking findingsReviewed Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
UX Review (Opus 5.5) — ✅ PASSUX-level review of UX-Verdict: PASS Nothing blocks: the screenshots show every changed control, and the blind reader read each one correctly. The Updates page still has a working-looking Update next to the refusal notice, the same problem this PR fixes on the detail page. Notes
[UX-REVIEWED] 86f7cda Description read[DESCRIPTION-READ] dfd09f93a4ff7eb4ceacfec501cd4fcf221636fceff17f341d722107c81844aa That sha256 names everything this verdict read: the PR description, plus the 8 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
GPT 6.1 Review — ✅ no blocking findingsGPT 6.1 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
abf2f89 to
d286801
Compare
|
First Principles review on abf2f89: agreed, the switch-sentence rewrite was outside the stated goal. d286801 puts |
d286801 to
59e586c
Compare
|
UX review on d286801. Accepted the block. A disabled Update that kept its blue fill still read as ready to press, so 59e586c hides the plain Update while the local-only notice is up. That covers both Update buttons and the overflow menu item, and the overflow menu only opens when it still has Uninstall. The now-unused update_held_for_repository_trust string is gone from all 13 locale files. AppDetailPageAutoAction.test.tsx asserts no Update button under the notice, and it fails when the hide is switched off. The new detail screenshot in the description shows the notice with Trust the repository and update as the only update action. The note about MORE_BELOW_MASK on the phone dialog body is about the dialog's existing scroll behaviour. This PR does not change it and it is outside the goal of #35913, so it stays out of this change. The take-back line now sits second, under the scope line, so it is in view without scrolling on a phone. |
…l to the trust step Follow-ups from the UX review of #30979 (issue #35913). - The detail page hides its plain Update while the local-only notice is up, so the notice's trust button is the only Update on screen. Pressing the plain Update only reran the refused update, and a disabled copy kept its blue fill and still read as ready to press. - Update All's repository-trust and chat-control consent lines carry an "Open <app> details" button per named app, shown only while that notice is on screen. A trust app's button reruns the update on the detail page; a consent app's button only opens the page, since its update landed. - A trust retry that meets the local-only refusal hands off to the detail page, as a first refusal does. - The consent dialog's take-back line renders second, under the scope line, on every path. On a 390x844 phone it ended 283 px below the footer on the switch dialog and 139 px below it on the enable dialog; it now ends above both.
…otice The detail page's two Update buttons now carry an `!updateHidden` guard, so the locations test's line filter found none. Accept the guard.
59e586c to
86f7cda
Compare
Code change approver recommendation: A person should decideWhy: the decider estimates a 10% chance that this change introduces a defect, within the 12% limit for approving on its own, but it is held at tier 2: 23 files, more than 20. Approve: this frontend-only follow-up does what it says, and each behaviour has a new test. On the detail page, the plain Update is hidden only while the local-only notice is on screen (all three Update spots are covered), and dismissing the notice brings it back. The Update All buttons show only while that exact notice is up, and the list is cleared at the start of each batch. A repository-trust app's button reruns the update on the detail page; a chat-control consent app's button only opens the page. A trust retry that meets the local-only refusal now resolves and opens the detail page. The trust dialog then closes normally and keeps the local grant the user approved; any other error is still re-thrown. The take-back line now renders once, second, on every dialog path. All 13 locales have the new string, and CI and the review lanes are green. Tier: 2, held for a person: 23 files, more than 20. An automated recommendation, not a review: a maintainer still decides. It approves on its own only when the estimated chance of a defect is at most 12% and the change has at most 400 changed lines. Decider v3c (217fe607) · commit |
KiroCrewCaptain
left a comment
There was a problem hiding this comment.
Approving. On this head the review read the whole frontend diff against main. Each of the four points in #35913 maps to a hunk. The retry after a local-only refusal hands off to the detail page and changes no grant. All three plain Update sites are gated on the notice that is shown. The Update All buttons are keyed on the exact notice text. The new key exists in all 13 locales. The 5 changed test files pass (147 tests, plus three shuffled runs), and reverting each production hunk turns a named test red.
Well done: every hide and every handoff button is tied to the notice that is actually on screen, so a dismissed or replaced notice brings Update back and drops the buttons, with no new lifecycle state.
Optional:
- The self-managed Update hides (
AppDetailPage.tsxaround lines 1489, 1502 and 1508) have no test, and the regex inapps.locations.test.ts:131makes!updateHidden &&optional. Aresources: 'app'variant of the local-only case, or a regex that requires the guard, would pin them. - The "Open details" button in the trust notice reruns the update through
updateApp. It lands on the right step, but a label like "Continue updating " would match what it does, and the UX and Design notes on this point still have no disposition. - Nothing pins
trustHandoff.message === displayErrorinDiscoverPage.tsx. One test that raises a second error after the batch and checks that the button is gone would cover it.
Problem / Motivation
Goal: After an update meets the local-only trust refusal, every App Store surface points the user at the one step that works, and the consent dialog shows its take-back line on a phone without scrolling.
#30979 added "Trust the repository and update" for an app installed from a local folder that a registry offers a newer version of. Four rough spots were left:
Why it matters
Anyone updating a folder-installed app sees two Update buttons and cannot tell which one works. Update All's notice names a step with no way to reach it. On a phone, the line that makes the consent safe to give is a scroll away, and the UX lane's reader declined there.
Not a goal
What changed (motivation → approach → change)
After the local-only refusal, the plain Update only repeated it, Update All named a step with nothing to press, and a phone hid the take-back line. Each surface below points the user at the step that works.
The Update All buttons appear only while that notice is the one on screen. A repository-trust app's button reruns the update on the detail page, which offers the trust. A chat-control consent line ("Review the permission in the app details") gets the same button, but it only opens the page, because that app's update already landed.
The plain Update is hidden instead of disabled. A disabled copy kept its blue fill, and the UX lane's reader still took it for a working button.
One new string,
open_app_details, is translated in all 13 locales. The dialog's own wording is unchanged.The app-kit spec names the hidden Update and the Update All buttons.
Backwards compatibility
Compatible: frontend only. No route, payload or stored data changes.
Reader: website/src/pages/AppDetailPage.tsx:updateHidden -- crew page -- while the notice is up, its "Trust the repository and update" button runs the same update with the trust it needs (
AppDetailPageAutoAction.test.tsx)Tests
AppDetailPageAutoAction.test.tsx: no plain Update is shown under the local-only notice, and the notice's button still runs the update.useAppUpdates.test.tsx: Update All hands the page every trust and consent app with whether its button reruns the update; a trust retry that meets the refusal opens the detail page.DiscoverPage.updatesTab.test.tsx: the notice shows "Open Notes details" for the refused app only, and it reruns the update; a consent app's button opens the page and starts no update.TrustAppModal.test.tsx: the take-back line renders once and comes before the capability list, on the switch path and on an ordinary update.Each new test fails with its fix reverted. The four files pass 5 runs in a row and 3 shuffled orders (seeds 11, 23, 47), 119 tests each time.
Proof: not provable -- frontend test (prove.py runs Python only)
Manual verification
A Playwright harness ran the built dashboard with dummy data. It pressed Update All, pressed "Open Field Notes details" and landed on the detail page. There it pressed Update, checked that no plain Update was left beside the notice, and opened the dialog from the notice. On a 390x844 phone the take-back line ends at 227 px, above the footer at 739 px. On main it ended 283 px below the footer on the switch dialog and 139 px below it on the enable dialog.
Screenshots / video
The phone dialogs, with the take-back line second:
The detail page with only the notice's Update, and Update All's notice with its button:
Switch dialog at desktop width
Related Issues
Closes #35913
Follows #30979.
Sibling sweep
Swept: Update All notice strings in
website/src/i18n/locales/en.jsonthat send the reader to an app's details (details). The chat-control consent line does too.Sibling: website/src/pages/apps/useAppUpdates.ts -- changed here
Swept: every
TrustAppModalpath that renderstrustAppModal.revocable, measured at 390x844. The enable and install dialogs had the take-back line below the fold too.Sibling: website/src/components/appstore/TrustAppModal.tsx -- changed here
Pattern harvest
Rule candidate: review-prompt
Pattern: a notice that tells the reader to open or press something must carry that control, and must not sit beside a control that repeats the refused action.
Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)