because after 4 years this has not be dealt with
Shit Punk Says
Oceans of Wisdom
19,158
matching drops
#1459158
2026-09-22 15:17
we have the same understanding of the situation and different conclusions
my point is that this step should probably not be in main right now
not before the above is done as you said
because my expectation is - every PR on main will fail for the time being
i mean effectively that is what you were doing anyway, right?
yes it is fine
but having and building proper E2E tests
the hard part is not having those
obviously most unit tests are mostly bullshit
it is something we will today
this is not some thing we will do at a "future date"
1/ we write pointless tests that we think are bullshit and then not enforce them
2/ either we think these unit tests add value in which case we should enforce them (and bring the repo and the tests into alignment with each other) or we don't think they add value and we should delete them
and of course broadly, the answer is not going to be
so you don't throw a hand grenade into its work
let me work with the bot that is working on this
I do not want you to randomly go changing the system
of which only 1 of them is "start enforcing the tests"
and since I have a different thread that is working for the last 2 days on implementing a large number of things
and presumably it is not the system you want, even though it is the system that was in place 2 days ago
is not the system I want
obviously the system of: we wrote stupid jest tests but we know to ignore them
that when finished may allow us to use bots at scale to build systems that touch crypto
I have spent quite a bit of time developing a more rigorous testing system
what I did this weekend is not simply TURN ON THE STUPID JEST TESTS
@[prxt0] and also all so that we are clear about expectations @[GelatoGenesis] @[simo] @[ragne]
wait
yes that someone is me
Short version: they mostly just started getting enforced.
What changed recently:
`App PR CI` was added to `main` on **June 20, 2026** in `#2801`.
The related-Jest command was fixed on **June 21, 2026** in `#2809`:
```diff
- ./bin/6529 run test:no-coverage -- --findRelatedTests ...
+ ./bin/6529 run test:no-coverage --findRelatedTests ...
```
That removed an extra `--`, so Jest now actually receives `--findRelatedTests`.
PR `#2556` then merged `main` on **June 22, 2026 at 05:30 UTC**, and immediately got its first `App PR CI` run. There were no earlier `App PR CI` runs on that PR branch to compare.
Why so many tests: the PR has **87 non-generated JS/TS changed files**, including shared auth/API/websocket/wave files. CI runs:
```bash
jest --findRelatedTests <all changed source/test files>
```
Jest’s dependency graph then selected **983 suites / 5536 tests**.
I also checked another fresh run, `Increase connected profile limit to five`: it only passed Jest `__tests__/services/auth.utils.test.ts constants/constants.ts`, but still selected **1156 suites / 6310 tests** and failed. That tells us the new gate is exposing broad/brittle existing test coverage, not that all those tests were newly rewritten.
So: not a mass recent test change. It is mainly new CI enforcement plus a broad dependency graph. Some PR source changes may create real failures, but the huge blast radius is the newly working related-Jest gate.
2/ point 2: i suspect what is starting to happen is that it has started actually enforcing the tests? I am in the process of finding out, but that is probably the case
so we can consider that part closed and go to point 2
1/ point 1: "i do not know what happened" -> 95% of the time with repo, aws cli, github access, codex can figure out what happened so one of my "this one weird trick developers don't want you to know" is to just "ask codex"
there are two separate points here right?
my extremely sophisticated prompt to find this out was to screenshot your question and add a '?"

The failing check is `App PR CI / Installed app checks`, run `#124`: [run 27931694159](https://github.com/6529-Collections/6529seize-frontend/actions/runs/27931694159).
Actual failure: `Run related Jest tests`.
Key line:
```text
Test Suites: 144 failed, 839 passed, 983 total
Tests: 392 failed, 5144 passed, 5536 total
```
So the Actions page is hiding a very large Jest failure set. It is not lint/typecheck: those passed.
Representative failures include:
```text
CreateDrop.test.tsx
Expected: "Drop submitted successfully"
Received: "Drop submitted."
CreateWave.test.tsx
Expected: "Please wait for image uploads to finish."
Received: "Wait for image uploads to finish."
MyStreamWaveTabsHeader.tsx:244
TypeError: Cannot read properties of undefined (reading 'handle')
```
There are also many fixture/test-env failures like `Cannot redefine property: matchMedia/location`.
My read: this PR touches broad shared auth/API/wave files, so CI’s `--findRelatedTests` pulled almost the whole Jest suite, exposing a lot of stale expectations and brittle fixtures. The immediate fix is not one failed app check; it is a cleanup pass on the related Jest failures, starting with updated copy expectations and wave fixtures missing `author.handle` / similar required fields.
@[prxt0] I know I am a broken record, but generally codex can answer questions like this, right?
once all of this is in place, we will be in a stronger position to scale
and then get the deployment trains working right
and then do the same for the other repos
2/ the bigger work has been going on all weekend on frontend which is putting in a more comprehensive testing framework that existed. bots have been running all weekend on this and are still going. we will get into it in detail when they are done
1/ I have added more review bots for us to see how they do:
a) a glm-5.2 cluster bot to see if this gives a different 'viewpoint' on issues
b) some specialized bots for backend, safe app, stream, etc
this is primarily in two buckets
the underlying infrastructure that will allow us to scale is having strong confidence in our testing regime
i want to put strong foundations in
so i spent the weekend working on testing/qa
gm
they are highly rated but you are not yet joined
they are not recommended for oyu
but that is not actually true
i want to say "recommended for you"
Highly Rated isn't exactly the right text
i think top only images @[simo] with a right arrow at the top right on that box to load the next 5 etc is ok, very very mild for now
yes you need more
not enough 🥵
24GB