JS optional arguments get corrupted

[robertbastian]

new icu.TimeFormatter(icu.Locale.fromString("de"), icu.DateTimeLength.Long, icu.TimePrecision.Minute, icu.DateTimeAlignment.Auto)
  .format(new icu.Time(12, 12, 12, 12)) // 12 Uhr

Changing the alignment to Column produces 12:12, but Auto should show the minutes as well.

[sffc]

Does not seem to reproduce via FFI in Rust. #6715

[robertbastian]

I've verified that this in an error in JS bindings, might be an ABI error even:

> new icu.TimeFormatter(icu.Locale.fromString("de"), icu.DateTimeLength.Long, icu.TimePrecision.Minute, icu.DateTimeAlignment.Auto)
  .format(new icu.Time(12, 12, 12, 12))
> diplomat-wasm.mjs:22 de Some(Long) Some(Hour) Some(Auto)

[robertbastian]

The first option also doesn't get converted correctly:

> new icu.TimeFormatter(icu.Locale.fromString("de"), icu.DateTimeLength.Short, icu.TimePrecision.Minute, icu.DateTimeAlignment.Column)
  .format(new icu.Time(12, 12, 12, 12))
> diplomat-wasm.mjs:22 de Some(Medium) Some(Minute) Some(Column)

[robertbastian]

I found a couple of things but I haven't been able to fix:

• optionToBufferForCalling doesn't return anything, should it?
• all optional arguments to the TimeFormatter constructor are written to the same buffer offset
• the struct allocated for the option does not seem to have space to hold the is_ok flag

@Manishearth @ambiguousname

[Manishearth]

optionToBufferForCalling doesn't return anything, should it?

Yes.

all optional arguments to the TimeFormatter constructor are written to the same buffer offset

It's four separate buffers, so that should be fine, right? Can you point out what you're noticing here in the code?

the struct allocated for the option does not seem to have space to hold the is_ok flag

Yeah, that's another bug.

We did a bunch of fixes to the option code earlier, I'm surprised this slipped through. optionToBufferForCalling is definitely something that's been iterated on a bunch.

[robertbastian]

It's four separate buffers

Oh yeah