Niklas Mischkulnig

Wouldn't this cause problems though in the general case because the order changes and therefore the specificity?

Maybe to force a new compositor layer? That probably doesn't happen with `translate(0, 0)` https://aerotwist.com/blog/on-translate3d-and-layer-creation-hacks/

I think I've first seen this on CI sometime in the last ~2 weeks. And no user appears to have actually run into this so far.

The failing benchmark job is unrelated, so all tests passed. But I'll report back if I see this error again in the future

``` 1) cache should support moving the project root: Invalid argument: MDB\_BAD\_RSLOT: Invalid reuse of reader locktable slot Error: Invalid argument: MDB_BAD_RSLOT: Invalid reuse of reader locktable slot at LMDBStore.getBinaryFast...

> does this test involve actually moving the location of the LMDB file/directory, possibly while it is in use? [It's definitely moved](https://github.com/parcel-bundler/parcel/blob/50893a7ec2b1490e6d0a2f5e1f9ab6a8fc57a113/packages/core/integration-tests/test/cache.js#L5883-L5910). There are no explicit calls to the "old"...

What do you propose we should do? Also, doesn't homebrew-core usually have the rather strict requirement that everything is compiled by Homebrew?

This is where the loading happens: https://github.com/parcel-bundler/parcel/blob/e294eafd9a49c056fb4b223c5ace5c0653428ede/packages/core/core/src/loadDotEnv.js#L10

> That is, assuming JavaScript regular expressions are fairly compatible with C++11's... 🤔 There is an ECMAScript std::regex flavour which is also the default if not set explicitly in the...

I'd say terser (and its config file) should be listed on https://v2.parceljs.org/languages/babel/