Martin Kustermann
Martin Kustermann
> That may be so, but it's a hack and can regress other workloads ... Checking go/lem: ad33dff4e9cd613defe294935beef1e296ed2975 regressed e.g. `IsolateFibonacci` by 100% in aot mode (i.e. it become twice...
May I ask why we make this an **optional** override? It means sometimes kernel ASTs know their target OS/Arch/... and sometimes not. This puts us in an awkward spot where...
Are we only looking into this for the purpose of reducing AOT code size? If so, then it seems to me the simplest solution would be to provide OS/Arch/... once...
> That's not a const @nate-thegrate. The issue specifically asks for a const solution to ensure tree shaking works properly (e.g. no Android code ends up in iOS build and...
> If you think these particular patterns are common and urgent to get in then I can look into it, but if this is more of a general issue then...
Users of this system will write `hook/build.dart` and `lib/foo.dart` and it's **very** confusing if the two don't use the same way to identify assets. And using `const DataAsset(package: 'foo', name:...
> using const DataAsset(package: 'foo', name: 'bar') is IMHO worse than const DataAsset('package:foo/bar'). > > I disagree. `const DataAsset(package: 'foo', name: 'bar')` is to me as a user clearly *...
> 'shader:flutter/circle' > > That's not allowed currently. Maybe we should? Or maybe we should clamp down our design and say we're never going to allow other schemes than package:?...
> I've cooked up some CLs to change the API in this package: > * https://github.com/dart-lang/native/pull/1536 > * https://github.com/dart-lang/native/pull/1538 > * https://github.com/dart-lang/native/pull/1539 We should set us up for a future,...