Repository navigation
Replies: 2 comments 2 replies
|
Hi @chemang It's Turbopack's PostCSS subprocess dying silently during HMR. The main process ends up blocked in poll() waiting on a child that's already gone + the CSS worker crashes, but Turbopack doesn't notice and sits there instead of erroring out. First clear everything and restart. Delete .next and node_modules/.cache, then run module.exports = {
plugins: {
'@tailwindcss/postcss': {},
},
}If you have any extra plugins in there, remove them and test. Plus temporarily comment out your custom styles in globals.css, leaving just the Tailwind imports. If the hang goes away, something in your CSS is crashing the worker. That's your repro. Don't forget to try the latest canary. Open an issue if you can isolate it. A minimal repro with the exact CSS that triggers the hang is something the Turbopack team will want to trace. The poll() blocked on a dead FD gives them a clear starting point. In the meantime, |
|
Hi, I've seen around some communications about v4's file watcher causing issues. This lines up with that. I'd say, make sure you are on latest tailwind to begin with, and also that you give Next.js 16 ago, for the most part, the migration should be straight forward. I believe you can run: |
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
I've been dealing with an issue for the past few days where next dev --turbo will randomly hang after a few HMR updates. The dev server becomes unresponsive, CPU drops to 0%, and no errors are logged anywhere. It's happening consistently enough that I've had to fall back to webpack, but I'd really like to use Turbopack for the speed.
I've been dealing with an issue for the past few days where next dev --turbo will randomly hang after a few HMR updates. The dev server becomes unresponsive, CPU drops to 0%, and no errors are logged anywhere. It's happening consistently enough that I've had to fall back to webpack, but I'd really like to use Turbopack for the speed.
My setup:
What I've tried so far:
The timing pattern is strange: it's not strictly reproducible on the Nth save. It tends to happen after I've been editing quickly for a while makes me think it could be related to some kind of internal queue filling up, or a race between HMR updates and the PostCSS reprocessing.
My questions:
Is there a known issue with Tailwind v4 and Turbopack where the CSS processing can deadlock under heavy HMR? I've seen mentions of PostCSS subprocess crashes in older Turbopack issues, but those seemed to produce errors, not silent hangs.
Is there a way to set a timeout on Turbopack's internal operations so that instead of hanging forever, it crashes with an error I can actually debug? Something like a TURBOPACK_TIMEOUT env var or a config option?
Are there any Turbopack-specific logs or diagnostics I can enable beyond NEXT_TURBOPACK_TRACING=1 that would show me what subprocess it's blocked on? I tried --reporter ndjson but that seems to be a pnpm thing, not Turbopack.
For anyone else who's hit this... did you find a workaround that's better than just disabling Turbopack entirely? Like splitting the CSS differently, or some Tailwind v4 config change that reduces the CSS processing overhead?
I'll attach logs if someone can point me to which log level or trace flag would be most useful. For now I'm stuck on webpack and the rebuild times are noticeably slower.
All reactions