A window snapped to half the screen (640-800 px) or a 1366x768 laptop
at 150 % (911 px) got the phone and tablet layout: bottom tab bar,
folders in a drawer, Settings as a phone list with a Back arrow. The
layout switched at the 64rem (1024 px) breakpoint.
Now the phone layout starts below 600 px. From 600 to 1024 px the app
is a compact desktop: app rail at the side, folder sidebar folded to
its icons (the existing collapsed state, opened per session with its
toggle), Settings with its two panes and a narrower page list. Below
840 px the split layout opens a message in the list's place, as the
focused layout does. This holds in the browser and the desktop app,
in the Pro shell (which now runs from 600 px) and under Mountain View.
- `hooks/use-media-query.ts`: classifyWidth, PHONE_BELOW 600,
FULL_DESKTOP 1024; the tablet class (`isTablet`) is gone, the store
has `isCompact` instead. Plugins still see "tablet" for it.
- `app/globals.css`: a `desk:` breakpoint at 37.5rem for the layout
classes, and a `tap:` variant for 44px rows on phones and touch
screens below 1024 px (was every window below 1024 px).
- Touch screens below 1024 px keep swiping list rows (was phones only).
- The boot skeletons, boot script, mail app, sidebar, reader, admin
shell and calendar, contacts and files follow the new widths.
Tests changed with the behaviour: the tablet cases in pro-gate,
mail-app-mv-composer-move, mail-app-compose-handoff,
mail-app-thread-expand-mark-read, mobile-tour-end and
thread-list-item-columns now run on a phone or a compact desktop, as
those widths no longer exist as a tablet. mail-app-short-window-tab-bar
is replaced by mail-app-compact-layout: a phone on its side is a
compact desktop with the rail at the side, so the short-window tab bar
rule is gone. Class and media-query assertions follow desk:/tap: and
37.5rem; the 32px-text test uses 1100 px, since 1440 px at 32px is now
a compact desktop.
The default layout showed only the newest message of a thread in the
reader and unfolded the rest as rows inside the list. The reader now
shows the conversation on every layout but a phone's, as Mountain View
does: older messages folded to one line each, unread ones unfolded, the
newest open. A conversation is one row in the list, with no chevron,
and `x` no longer unfolds it. "Show emails as individual messages"
keeps the single message.
Two existing tests changed their setup, not their assertions: the
attachment download tests in `blob-routing.test.tsx` turn threading off
so the conversation fetch doesn't take the requests they count, and the
`x` expansion tests now run on Mountain View on a tablet, the one place
where the list still unfolds a conversation.
The selection bar showed Archive, Delete and Tag, with Move only in its
more menu and both read buttons only on wide bars. The reading pane said
"No conversation selected" next to a selection, and the first-run tips
took a third of the list.
- The bar has Archive, Delete, Move, one read/unread button ("Mark as
read" while anything selected is unread) and Tag, then the rest.
- With a selection the empty reading pane shows "5 conversations
selected" and the same buttons.
- The first-run tips sit in the empty reading pane where it shows beside
the list; phones and layouts without that pane keep them above the list.
- A bulk move's toast names the folder.
The formatter cache test expected the numeric date for every older mail; earlier this year it now reads "Oct 24" unless a date order is picked in the region settings, so the test expects that for those dates.
The badge test said a differing sender name gets "via <identity>"; that badge is gone, so it now checks that nothing is shown and that the sub-address tag still is.
The toolbar measure test counted only item width reads; a pass that fits now reads no item widths, so it counts the toolbar's own scrollWidth reads too.
Two sidebar tests asserted the old unread-count button (it filtered the
folder to unread without opening it, in a 24px target); they now check
that the count is plain text and opens the folder.
The default layout's switcher listed accounts without their unread, so an
existing test asserted that (account-switcher-mountain-view, 'unread-free');
it now expects each account's count. The idle-outside-Pro hook test now
uses one account, since a second connected account loads in every layout.
The hand-off parked a password before its question, for up to five
minutes, and on a shared computer another local user could ask for
someone's sign-in with a link of their own. Now nothing leaves the page
before the yes, and the instance has a switch for such machines.
- `/check` tests a password and keeps nothing, `/park` runs on the yes and
hands over tokens wherever the server allows, a password as a last
resort
- `/discard` and a timer take out what SSO has to park first, and a no
revokes the refresh token, and the shared store's TTL is the real expiry
- `DESKTOP_BROWSER_SIGN_IN`, default on, closes the routes, the page's
hand-off and the app's button; the question says which device asked
- tokens from a typed password stay a password account under the name
typed (#1432)
- redeem refuses a server of the user's own when the device's rule says
no
- the park budget is reserved before going upstream, and only failures
count
- "Not now" ends the flow in the app at once, and the question names the
account
`desktop-handoff-routes.test.ts` now tests the two-step flow against the
real budgets. `sso-start-route.test.ts`'s config mock answers the new
setting, and the SSO test's expected answer has the account in it.
The first version of quick compose skipped `onBeforeCompose`, a composer
moved back could end up open in both windows, a reply moved back under
Pro flagged the selected message, and a crashed pop-out came back blank.
- A new message in a pop-out runs `onBeforeCompose` with the same result
as the main window: a plugin that stops it closes the window.
- The way back has two steps, so a pop-out that gave up can't leave a
second owner. A handoff behind a plugin's gate makes it say busy.
- A Pro tab made from a composer moved back answers its own message.
- Each save writes `draft`, `account` and `owner` into the pop-out's
address, which a reload after a crash starts from. What was typed since
the last save, up to the 2 s autosave delay, is still lost.
- The recorder refuses keys that type a character on this keyboard, and
what the shell refuses from a page, and says why.
The desktop login page gets "Sign in with your browser…" as its main
button. On an instance it opens the instance's real login page in the
system browser, with a PKCE challenge. The page checks the password or
runs SSO, parks the sign-in under a one-time code, asks once, and sends
the code to the app's loopback listener. This browser gets no session
out of it, and the app's own page redeems the code, so SameSite=Strict
is fine. Standalone runs the mail server's OAuth against the listener,
with a page from main for a server with no OAuth at all.
- `/api/auth/desktop-handoff/park` and `redeem`, and `sso/complete` parks
tokens the same way for a loopback target
- `auth.claim()`: the calls are taken once per document, in the first
render, in `WindowControlsHost`
- `BrowserSignIn` policy, default on, with its ADMX and samples
- strings in all 30 languages, and the old button's are gone
`runDesktopBrowserSignIn`'s tests changed because the code no longer goes
through the callback page's address: it's redeemed where it arrives.
Linus Rath