[khagele]

A visitor asked in Discord where to make an account for the map and could not find one. There is no answer to find: the login card offers login and nothing else, and the map cannot register anyone anyway, since /api/auth/register rejects a body without a companion_pubkey (server/internal/httpapi/auth.go). Only the RX webapp has one.

That gap turned out to be a positioning one. The RX webapp is the product: you pair a companion to it and map the mesh, and the map is where every mapper's receptions meet. Both sites led with the map instead, which sends a first-time visitor to a login screen for data they have no part in yet.

The map

• #rx-cta, a Start mapping link in the bar next to Log in, opening rx.mesh-hunter.eu.
• The login card gains a footer line: no account yet, register in the RX webapp, it links the companion your phone is paired to, an admin verifies you as a member afterwards. Inside the <form>, so the focus trap includes it.
• The walkthrough explains the button, is anchored to it, and its panel now says where mapping happens.
• The guest notice goes back to "Log in to see more" (see below).

The landing

• Nav lists the RX webapp before the map, on all four pages.
• Hero leads with Start mapping and names the companion pairing; "See the map" is secondary.
• The RX webapp section comes before the map section, and a new "Get an account" section carries the four steps through to admin member verification, out of the small print at the foot of the page.
• Mapper and hunter stay both words in the copy: mapping is what the radio does for you, hunting is going after one node. Role names are untouched.

Two measurements, because #bar wraps by content width

• Naming the RX webapp in the guest notice took it from 703px to 800px, which pushed the SF counts onto a row of their own at 1440px. The notice is back to its old wording; the button beside it carries that message now.
• #rx-cta costs no bar row at 1024/1280/1440. At 375px its long label does, so it shortens to "Map it" below the 640px breakpoint: that row holds ~80px, and "Start mapping" is 109px.

Checks

• npx vitest run in web/ (413), app/ (716) and landing/ (35); npx playwright test in web/ (159); eslint clean.
• New tests were red first, and the reworded guest-notice assertion was mutation-checked by putting the long copy back.
• Verified in the browser at 375/1024/1280/1440 in both themes: bar height unchanged with the button present, the login card footer renders, the fragment link to Get an account clears the sticky header.

Closes #490

[khagele]

Rebased onto master for #497. The only conflict was changelog.json, where #497 added an entry of the same date; both are kept, both copies byte-identical.

One thing that changed with the rebase and is worth stating rather than leaving to be noticed: #497 filled the toolbar up, so #rx-cta now costs a bar row where it cost none before. Measured after the rebase:

viewport	without the button	with it
1440	83 px	119 px
1280	119 px	156 px

Short label and icon-only were both measured too: at 1440 the bar wraps either way, so the row is the price of having the control at all, not of its wording.

Kept, deliberately. The map is downstream of the RX webapp, and the one visitor this whole PR exists for is the one who has never mapped — 36 px of map height is what that costs on desktop. The mobile breakpoint still shortens the label to "Map it", where the bar has the room.

web 423, landing 35, playwright 166, eslint clean.

[efiten]

Reviewed the code; the product call is the maintainer's. Everything I checked holds up, and the one thing I would ask for is a re-run rather than a change.

The green here is stale, and it matters for this PR in particular

CI last ran at 04:56; #528 merged at 09:39 and added a Sender id toggle to #bar (76px). This PR adds #rx-cta (106px) plus a seventh callout. That is the same density that broke three onboarding.spec.js tests on the runner yesterday while passing on Windows, so the two changes land on exactly the shared resource.

I merged current master into this branch locally and ran it. It holds:

bar height   1024: 161   1280: 125   1440: 86
widths       rx-cta 106   f-idclass-wrap 77   f-types 882
onboarding   17 passed
web          479 unit + 182 playwright passed
landing      35 passed

Local Windows metrics are the narrow ones though, and the runner's are what caught it last time, so please merge master in and let CI say so rather than taking my machine's word for it.

(landing.test.js failed for me at first on a missing fast-xml-parser -- my checkout, not this branch. Clean after npm ci.)

Checked, and each one holds

• "Inside the <form>, so the focus trap includes it." Accurate: login.js:53 is trapFocus($('login-form')), so the trap really is scoped to the form, and login.test.js slicing between id="login-form" and </form> pins precisely the thing the claim rests on rather than merely the link's presence.
• The guest notice. The diff adds a longer wording and then takes it back out, with the measurement recorded (972px against a row that holds ~800px). Net effect on the notice is nothing, which is the right outcome; worth saying out loud that the PR body's "goes back to" is the whole story and not a leftover.
• The responsive label. .rx-cta-long / .rx-cta-short swap on display, which drops the hidden one out of the accessibility tree too, so a screen reader gets one label rather than "Start mappingMap it". No aria-hidden needed.
• One callout per control. Splitting wb-co-mapping out rather than anchoring one box to the union of rx-cta and auth-btn is right, and the comment gives the reason plus the width it was found at.
• Both link targets resolve: rx.mesh-hunter.eu and mesh-hunter.eu return 200, rel="noopener" present on every new target="_blank".

Not a blocker, one thing left unsaid

The login card now tells a visitor to register in the RX webapp and that "an admin verifies you as a member afterwards". Nothing tells them how long that takes or how they would know it happened. Not this PR's job to build, but it is the next dead end on the same path, and worth an issue rather than being discovered by the next person who asks in Discord.

[efiten]

Filed the open end I mentioned as #530, so it is a decision rather than a discovery. Not a blocker for this PR.

[efiten]

Approved. The three calls in here are the maintainer's, and all three are accepted (2026-08-26): the RX webapp leading over the map, "mapper" alongside "hunter" in the copy, and the account path spelled out.

Re-verified against current master, which has moved a long way since this branch's CI ran at 10:11 (#477, #500 and three releases). Merged master in locally and ran the lot:

web       510 unit, 193 playwright
landing    35

Onboarding included, so the extra bar control still costs no callout.

One thing said plainly rather than hidden: the first of my four playwright runs reported 1 failed / 192 passed, and I lost the output before identifying which test it was. The three runs after it were clean at 193. So it looks like a one-off flake rather than a fault in this branch, but CI runs with --fail-on-flaky-tests and is the arbiter, not my laptop. If it turns up there, it is that, not something new.

The code review from my earlier pass stands unchanged: the focus-trap claim holds (login.js:53 traps on #login-form, and the test slices exactly that), the guest-notice wording is added and then taken back out with the 972px measurement recorded, the responsive label swaps on display so a screen reader gets one label, and both link targets return 200 with rel="noopener".

The open end this leaves, that nothing tells a hunter when the verification actually comes through, is closed separately in #530 and shipped.