just up works and ends on a clear success line. Getting there, a first-time user reads eleven lines of warning first.
What a clean run actually prints
Seven grey [fold:run] notes:
[fold:run] observability/monitoring.ts — "serviceMonitor" is not foldable: 10:12 - unresolved
identifier: seams (2:42 - "backupRetentionDays" is not foldable: 38:21 - function call as a
value is not foldable: resolveTier(...))
[fold:run] params.ts — "backupRetentionDays" is not foldable: ...
[fold:run] secrets/infisical.ts — "infisicalSecret" is not foldable: ...
then four yellow warnings:
warning: Container "dump" in CronJob has no explicit imagePullPolicy
warning: Container "dump" in CronJob does not drop ALL capabilities
warning: Deployment "fountain" has 1 replica(s) — consider at least 2 for high availability
warning: 4 post-synth finding(s) suppressed via lint.rules (severity "off")
then the thing you actually wanted:
✓ /health answered
fountain is up. Reach it with: just forward → http://localhost:4000
Why it is worth fixing rather than ignoring
Every one of those lines is either expected or deliberate, and none of them says so. A first-time user has no way to distinguish "this is fine" from "this is the reason it will not work", so the honest reading of a successful deploy is that something went wrong.
It also trains people to ignore the warnings, which is worse than not printing them — the two CronJob findings are real and unaddressed, and they are now camouflaged by the noise around them.
The fold notes have one root cause
All seven trace to the same line: resolveTier(...) is a function call, so backupRetentionDays is unfoldable, so everything downstream of it is too. It is one fact reported seven times from seven call sites.
Whether that is worth fixing structurally (a const table instead of a call, at some cost to readability) or worth suppressing on the up path is a real choice. What is not defensible is printing it seven times on every successful deploy.
The 1 replica warning is wrong here, specifically
light is single-replica by design, and the README explains why more than one replica needs an Erlang cluster. Warning about it on every run of the tier whose whole point is one replica is noise that the repo has already answered.
Options
- Quiet the build inside
just up and keep the full output on just build, so the loud version is one command away when something is actually wrong.
- Suppress the replica warning at tiers that are single-replica by design, the way
WK8201 and WK8203 are already switched off in chant.config.ts with reasons.
- Fix the two real CronJob findings so they stop being noise: an explicit
imagePullPolicy and a drop: ["ALL"] on the dump container.
- Address the fold chain at its root so the seven notes become zero.
The last two are worth doing on their own merits regardless of what happens to the output.
just upworks and ends on a clear success line. Getting there, a first-time user reads eleven lines of warning first.What a clean run actually prints
Seven grey
[fold:run]notes:then four yellow warnings:
then the thing you actually wanted:
Why it is worth fixing rather than ignoring
Every one of those lines is either expected or deliberate, and none of them says so. A first-time user has no way to distinguish "this is fine" from "this is the reason it will not work", so the honest reading of a successful deploy is that something went wrong.
It also trains people to ignore the warnings, which is worse than not printing them — the two CronJob findings are real and unaddressed, and they are now camouflaged by the noise around them.
The fold notes have one root cause
All seven trace to the same line:
resolveTier(...)is a function call, sobackupRetentionDaysis unfoldable, so everything downstream of it is too. It is one fact reported seven times from seven call sites.Whether that is worth fixing structurally (a const table instead of a call, at some cost to readability) or worth suppressing on the
uppath is a real choice. What is not defensible is printing it seven times on every successful deploy.The
1 replicawarning is wrong here, specificallylightis single-replica by design, and the README explains why more than one replica needs an Erlang cluster. Warning about it on every run of the tier whose whole point is one replica is noise that the repo has already answered.Options
just upand keep the full output onjust build, so the loud version is one command away when something is actually wrong.WK8201andWK8203are already switched off inchant.config.tswith reasons.imagePullPolicyand adrop: ["ALL"]on the dump container.The last two are worth doing on their own merits regardless of what happens to the output.