fix(ci): stop push runs from preferring stale mainnet secret#432
Conversation
Greptile SummaryReviewed PR #432 which fixes CI by hardcoding Confidence Score: 5/5Safe to merge — the one-line change is targeted, low-risk, and directly addresses the confirmed CI regression. The diff is a single-line replacement that removes a secret reference in favour of a hardcoded public URL. The logic is straightforward, the root cause is well-evidenced in the PR description, and the only remaining item (stale comment on line 79) is a non-blocking style nit. No files require special attention. Important Files Changed
|
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## dev/v4-fin #432 +/- ##
=============================================
Coverage ? 58.83%
=============================================
Files ? 12
Lines ? 1834
Branches ? 0
=============================================
Hits ? 1079
Misses ? 755
Partials ? 0 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
|
Confirmed the mainnet regression is fixed here.
I am not broadening this PR to paper over that unrelated Arbitrum failure. This PR does the narrow thing it was opened for: stop push runs from preferring the stale mainnet secret. |
|
Addressed Greptile's P2 nit (3f73505): updated stale comment on line 79 to reflect that all provider env vars are now hardcoded public endpoints (no longer "use secret if available"). Everything else is unchanged — Greptile 5/5, mainnet CI green, Arbitrum failure is pre-existing and unrelated. |
Summary
secrets.MAINNET_PROVIDERhttps://ethereum.publicnode.comfor mainnet CI forksWhy
PR #431 fixed the fallback URL, but the post-merge push run on
dev/v4-finstill consumed the repo secret on push jobs. That secret is a stale Infura URL, so Ganache still failed withToo Many Requests.Evidence:
PROVIDER_MAINNET: https://ethereum.publicnode.comand passed the mainnet jobsb2eb3a0showedPROVIDER_MAINNET: ***and failed withInvalid response from fork provider: {"code":-32005,"message":"Too Many Requests"...see https://infura.io/dashboard}This makes push CI behave differently from the PR run, which is dumb. CI should use the known-good public endpoint directly until the secret is updated or removed.
Notes
test (v3, arbitrum)failure is separate and was already present after fix(ci): replace rate-limited Infura with public RPC #431.