Skip to content

A coming-up Mate's page hands over as soon as the Mate answers, whatever the browser kept - #54

Merged
fxck merged 3 commits into
mainfrom
fix/arrival-handover
Oct 1, 2026
Merged

fxck merged 3 commits into
mainfrom
fix/arrival-handover

Conversation

@fxck

@fxck fxck commented Oct 1, 2026

Copy link
Copy Markdown
Member

In the browser that created a Mate, the arrival page could stay on "coming up · Almost there" long after the Mate was up and signed in, even through a reload. It was seen live on 2026-10-01: an hour in one browser, and 17 minutes in another.

Causes and fixes:

  • A leftover setup record won. mateComing checked "coming" before "linked", so a setup record still held in this browser kept "coming up" even with the Mate's environment registered. Now a Mate whose conversation can open always hands over.
  • Setup records weren't cleared. A setup record ended only when the install that wrote its environment record also promoted it. Now the environments stage ends the setup of every project with a record, at start and whenever the records change.
  • The Mate on screen might never connect. Only 12 Mates auto-connect, and the creating browser knew 21. Now the Mate whose page is open always connects, past that limit.
  • A failed connect never retried. The page connected once, by origin, and a refusal was never retried. Now it connects the target that the environment machine holds, and the machine retries on its own ladder.

Tests (RED → GREEN):

  • mateComing.test.ts, 2 cases;
  • accountRuntime.test.ts, a stage-start case;
  • autoConnect.test.ts, 2 cases;
  • mateConnectKey, 4 cases;
  • a page test of the Mate going on screen and off it.

🤖 Generated with Claude Code

fxck added 3 commits October 1, 2026 12:17
… coming says

A Mate's own view showed its coming first and its link only after: a browser that still
held the Mate's birth kept saying it was coming up after its environment was registered
and the Mate answered (a live run: an hour, through reloads, while another browser of the
same person worked in its conversation). Linked or connected now hands over first.
A birth ended only on the install that wrote its Mate's record; a browser that held the
record and still the birth kept the birth for good, and with it the Mate's coming. The
environments stage now ends the birth of every project its records name, when it starts
and whenever the records change, each once; ending one that is not held does nothing.
… and a failed connect is tried again

A browser with more Mates registered than auto-connect's ceiling never wanted a new one:
the ceiling held it back even while its own view was open. The Mate on screen is now
wanted first and past the ceiling, which is for Mates not on screen; its view puts it on
screen while it stands. And the view connected a Mate whose listing had not caught up by
its origin, which was turned away once and never asked again while the view stayed; it
now always connects a target its machine holds, the listed row's where no birth names
the container, so a connect that fails is the machine's to try again on its own ladder.
@fxck
fxck merged commit 2d8eb3f into main Oct 1, 2026
11 checks passed
fxck added a commit that referenced this pull request Oct 1, 2026
- #50 A person's own new project never waits on the path for someone else's
- #51 A docked operation offers to open only when opening adds something
- #52 A change's review reads the change itself, and rows name their repository when a Mate has two
- #54 A coming-up Mate's page hands over as soon as the Mate answers, whatever the browser kept
- #53 The account's line: Try now shows it tries, names what isn't answering, and one org's trouble holds no other
- #56 The composer's top lists every change that waits for review
- #55 A message echoed in the run card shows its words and its pictures
- #57 One stalled subscription retries alone instead of replacing the organization's socket
- #58 Docs: what the hours after 0.11.74 measured, and the fixes they took
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant