Fix Browser CI #3369
Fix Browser CI #3369
Conversation
acolytec3
force-pushed
the
vitest-update
branch
from
d06d988
to
4a57122
Compare
| import { nodePolyfills } from 'vite-plugin-node-polyfills' | ||
| import wasm from 'vite-plugin-wasm' | ||
|
|
||
| const config = defineConfig({ |
There was a problem hiding this comment.
| const config = defineConfig({ | |
| // https://vitest.dev/config/ | |
| const config = defineConfig({ |
There was a problem hiding this comment.
@roninjin10 do you have any insights on why an arbitrary browser test script seems to always fail on CI? All of our browser tests pass locally but when they run on CI, we get these failed to resolve dynamically imported module... errors. I've tried all manner of config settings trying to remove parallel test runs and isolated environments, etc, with no success. At this point, I haven't come up with a reproducible "minimal" build to point the vitest devs at so have hesitated to open an issue there (though may do so soon)
There was a problem hiding this comment.
@acolytec3 try setting VITEST_BROWSER_DEBUG=true environment variable
There was a problem hiding this comment.
Btw if the debug logging doesn't provide any insight let me know I'd be happy to look at this
There was a problem hiding this comment.
It just keeps doing this sort of thing - https://github.com/ethereumjs/ethereumjs-monorepo/actions/runs/8943487337/job/24568309037#step:9:74
Debug logging doesn't seem to tell me anything more. It's not always the tx test run that fails either. I can comment out tx and another one will fail the same way (or I can tell vitest to skip the one test file that's not being loaded and it will still fail on tx with a different file having no tests fail). Very weird.
There was a problem hiding this comment.
It seems to have started after vitest v1.5.1 was released
There was a problem hiding this comment.
If you don't have a quick instinct on it, I'll just open an issue on vitest and see if they're willing to look at our vitest config since I don't know how else to reproduce it.
There was a problem hiding this comment.
. It's not always the tx test run that fails either. I can comment out tx and another one will fail the same way (or I can tell vitest to skip the one test file that's not being loaded and it will still fail on tx with a different file having no tests fail). Very weird.
oh that is really wierd and definitely smells like an internal race condition bug in vitest
There was a problem hiding this comment.
That was my instinct as well. Will report it and see if they're willing to help.
|
Have restarted CI since blockchain and state tests were failing (a bit unexpectedly + couldn't get anything out of the logs) |
| - run: npm run test:browser -w=@ethereumjs/evm | ||
| - run: npm run test:browser -w=@ethereumjs/vm | ||
| - run: npm run test:browser -w=@ethereumjs/verkle | ||
| - run: npm run test:browser --workspaces --if-present |
| @@ -1,51 +0,0 @@ | |||
| name: E2E Hardhat Tests | |||
There was a problem hiding this comment.
Hardhat out, yeah, guess that makes sense.
Not so realistic we'll give this so much attention again in the future.
| exclude: [ | ||
| ...configDefaults.exclude, | ||
| // readDirSync method not provided fs mock for vite | ||
| 'test/precompiles/eip-2537-bls.spec.ts', |
There was a problem hiding this comment.
Guess this can be removed again (?) after BLS merge
There was a problem hiding this comment.
Update: just tested and realized it's another thing
There was a problem hiding this comment.
Yep, there are probably several of these where the mock that vitest injects a polyfill that's not fully mapping to the node builtin we're referencing.
Fixes browser tests