`String.codeUnits` behaving differently in dart2js vs ddc about sdk

[lrhn]

How do they behave differently?

If I check the the output of the second print (dart compile js using Dart SDK 3.5.0-43.0.dev for Dart2JS and dartpad.dev main channel, 3.4.0-190.0.dev, for DDC), they give the same numbers, so it's the same string and .codeUnits behaves the same.

Any difference would then be in how that string, containing weird code points, is displayed.
How is the code compiled and run, and how is the output displayed? (And how does it differ?)

(I have had problems with non-ASCII string literals before, when compiled into google3, because the source encoding wasn't respected. This CL failed to land because one of our web compilers generated a non-ASCII JavaScript string literal as UTF-8, and some internal processing took it as Latin-1, IIRC. Could be related to that.)

[rakudrama]

A test created from the example passes on all configurations: https://dart-review.googlesource.com/c/sdk/+/362700

[vsmenon]

Internal thread here: https://chat.google.com/room/AAAAS7MyFlc/6jkbFmEmEPo

Seems like something changed recently for that internal customer.

[lrhn]

So it seems the bug is in google3, not in the compilers.

[rakudrama]

I think the test I uploaded demonstrates that there is no discrepancy in behaviour of the code generated by dart2js and DDC, and the problem is elsewhere, either serving with the wrong charset encoding, or processing the files in some way that 'evaluates' the \udba1 escape.

Can we close this as not reproducable?

[vsmenon]

Thanks, @rakudrama !