Skip to content

docs(#17): the relay host runs main, at 2ea1e15 - #99

Merged
Lexirieru merged 6 commits into
mainfrom
docs/17-relay-redeploy-main
Oct 30, 2026
Merged

Lexirieru merged 6 commits into
mainfrom
docs/17-relay-redeploy-main

Conversation

@Lin1er

@Lin1er Lin1er commented Oct 25, 2026

Copy link
Copy Markdown
Collaborator

The relay host now runs main. This records it in docs/deployments.md, under the #16 / #17 section whose "the tree is not main" bullet it retires.

Docs only, no code.

The host was fast-forwarded from 29d6719 and restarted. The only
runtime change is #92's wording; health, config, ready and CORS were
checked from outside afterwards.
@Lin1er
Lin1er requested a review from Lexirieru October 25, 2026 14:09
@vercel

vercel Bot commented Oct 25, 2026 •

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated
forepay Ready Ready Preview Oct 30, 2026 1:47pm UTC
forepay-vue5 Ready Ready Preview Oct 30, 2026 1:47pm UTC

@Lexirieru Lexirieru left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@Lin1er — holding this one, because the sentence it records is no longer true, and docs/deployments.md is the file we trust when something breaks.

What the host answers now

I probed the live relay from outside, with a valid G… borrower and a deliberately invalid token, to see which provider it routes to:

POST https://api-forepay.jameshub.fun/proofs
{"source":"admob","borrower":"GA2CD6PM…WWH2","accessToken":"not-a-real-token"}

202  {"id":"b7e571d3…","stage":"checking"}
->   {"stage":"failed","error":{"code":"token","message":
     "Google refused the access token (401). It is missing, expired (they last an hour)
      or lacks https://www.googleapis.com/auth/admob.readonly."}}

The host names admob.readonly. That string does not exist anywhere in be/ on main:

$ git grep -n "admob.readonly" origin/main -- be/
(no matches)

So the host is running #100's AdMob provider, and #100 is still a draft. The host is ahead of main, not on it. 03da644 is where it was fast-forwarded to; it is not where it is.

Everything else in this PR's body I checked and it holds — /health 200, /config recorded: true, relay: true, advance CCQFL2Q6…I2HS, and CORS answering for https://app.forepay.xyz and nothing else. The AdSense path on the host is main's. It is the one bullet about the tree that has gone stale.

Why I am not merging it as written

This is the #102 shape again, in the records rather than the code. #102 survived two 7/7 GREEN rehearsals because the fixture and the contract were both derived from the same wrong assumption, and nothing was pinned to reality. A deployments.md that says the host runs main at 03da644, while the host actually runs an unmerged branch, is the same trap: the next person to debug a relay failure will read that line, diff main, and find nothing wrong.

Not a criticism of running it on the host — that is how you got a real AdMob proof from the production attestor, which is the strongest evidence this project has. It just has to be written down as what it is.

What would make this mergeable

Either is fine, your call:

  1. Record the truth. The host runs main at 03da644 plus feat/admob-relay at <sha>, with one line saying the AdMob provider is on the host ahead of its PR, and that the tree returns to main when #100 merges.
  2. Merge #100 first, redeploy, then this PR records main at the new sha and is simply correct.

If #100 is close, 2 is cleaner and I would rather review #100 today than carry a known-stale record through the upgrade. #107 is merged (354d9e0) and #104 with it (f415506), so main's be/ is now what the host needs for the upgrade — which makes getting the host onto a recorded main worth doing properly before execute_upgrade, not after.

One thing I should own: my probe started a real job on the live relay under the deployer account, which held that borrower's job slot for a few seconds. It failed at the token, as intended, and touched nothing on chain.

@Lin1er Lin1er changed the title docs(#17): the relay host runs main, at 03da644 docs(#17): the relay host runs main, at 2ea1e15 Oct 30, 2026
@Lin1er

Lin1er commented Oct 30, 2026

Copy link
Copy Markdown
Collaborator Author

@Lexirieru, done: the host is back on main at 2ea1e15 (fast-forwarded, not detached), deployed 2026-10-30 13:45 UTC, with checks from outside in the record. The title now says that. Its history stays in the doc: 03da644 (2026-10-25), then feat/admob-relay at 3102f64, 1095ba6 and 8097128, each listed at the time, until #100 merged. Your point stands: for those days the PR title described a state the host was no longer in. The body notes covered it, but the title is what people read.

@Lexirieru Lexirieru left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Approving. You did more than I asked: rather than correct one line, you reconstructed the whole timeline with each host state dated, including the two days it ran a detached HEAD. That is better than a record that was never wrong, because now someone debugging a relay failure on 2026-10-27 can see what the host actually was that day.

The entry I was least expecting is the one that matters most:

The host is on a detached HEAD, and goes back to main when #100 merges.

Written down at the time, with the reason. That is the opposite of the failure I was worried about.

Checked here

What Result
check_links.py docs/deployments.md on your head every link resolves, contract verified
/health {"status":"ok"}
/config recorded: true, relay: true, advance CCQFL2Q6…I2HS
The AdMob path on the host probed with a bad token: 202, then failed naming admob.readonly

One thing I could not verify from outside, so I am not claiming it: #107 being on the host. /advances for the deployer returns [], because production has never paid an advance — claim_template() is still null there — so there is no row whose source field I could look at. Your evidence for it is sound (354d9e0 is an ancestor of 8097128, and 2ea1e15 is main), and it is your checkout, but it rests on your record rather than on something I observed. Saying which half of a claim I checked seems worth the extra sentence given how this week has gone.

One nit, not worth a push: the PR body still opens with the 29d6719 → 03da644 fast-forward as though it were the whole story. The title and the document are right, and the document is the record, so I would leave it.

Where this lands in the upgrade order

This was step 1, and it is done before step 2 rather than after, which was the point. Since you redeployed, main moved twice more — ee05add and a2150cb — both docs only, no be/ change, so the host does not need another restart for them.

v0.5.0 is uploaded and proposed on both contracts. Wasm ff6a16e988d8c31f27695efff60172764eefae9b19e2b95914c010c509e4f22e, from the release artifact. Executable 2026-10-31 13:49:10 UTC on production and 13:49:20 on dev. Details and transaction hashes are in docs/deployments.md at a2150cb and on #102.

@Lexirieru
Lexirieru merged commit f582fd2 into main Oct 30, 2026
9 checks passed
@Lexirieru
Lexirieru deleted the docs/17-relay-redeploy-main branch October 30, 2026 13:53
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.

2 participants