[WIP] Add __traits(vtblSymbol) and __traits(interfaceSymbol), as __SymbolSlice

[dkorpel]

Illustration for #23991

[TurkeyMan]

Shouldn't this be immutable too? and if we're here, why not just make the instance immutable?

[TurkeyMan]

Oh, I have another one with a bunch of edits going too... should I abandon mine?
I'm not certain this is better...

[github-actions]

DMD perf check

Metric	Base	PR	Δ
compile hello.d (instr)	210.3 M	210.7 M	+0.215%
compile hello.d -O -release (instr)	227.9 M	228.6 M	+0.323%
dmd binary size (stripped)	8.15 MB	8.14 MB	-0.14%

Breakdown — compile hello.d

Phase (wall, self time)	Base	PR	Δ
sema_other	13.9 ms	14.1 ms	+1.85%
parse	34.3 ms	34.0 ms	-0.66%
sema1	10.5 ms	10.6 ms	+1.66%
sema3	5.5 ms	5.6 ms	+1.97%
codegen	2.0 ms	2.1 ms	+3.79%

All measurements

Metric	Base	PR	Δ
compile hello.d (instr)	210.3 M	210.7 M	+0.215%
compile hello.d -O -release (instr)	227.9 M	228.6 M	+0.323%
compile Phobos (instr)	4,891.1 M	4,894.7 M	+0.075%
compile Phobos codegen (instr)	1,426.1 M	1,426.8 M	+0.050%
compile vibe.d (instr)	14,257.4 M	14,269.3 M	+0.083%
dmd binary size (stripped)	8.15 MB	8.14 MB	-0.14%
hello binary size (stripped)	0.72 MB	0.72 MB	0.00%
peak RSS (compile hello.d)	43.89 MB	44.11 MB	+0.51%
peak RSS (compile Phobos)	605.0 MB	603.7 MB	-0.22%
peak RSS (compile vibe.d)	1858 MB	1859 MB	+0.09%
page faults (compile hello.d)	8,894	8,915	+0.24%
page faults (compile Phobos)	151,174	150,834	-0.22%
page faults (compile vibe.d)	465,192	464,532	-0.14%
compile dmd itself (wall)	12.1 s	12.2 s	+0.73%
compile hello.d (wall)	66.1 ms	66.5 ms	+0.59%
compile Phobos (wall)	1,473 ms	1,471 ms	-0.14%

_{0dbd861 vs merge-base 92ac966 · about these metrics}

[dkorpel]

Oh, I have another one with a bunch of edits going too... should I abandon mine? I'm not certain this is better...

I don't know if it's better either, I just wanted to test if the new determineVtbl code addition is redundant and I didn't want to push directly to your branch. Feel free to incorporate changes and I'll close this.

[TurkeyMan]

I think determineVtbl was redundant; my Claude decided the reasons he did it was to facilitate some additional use cases, which yours here rejects as a forward reference.

Like, I think your one here will fail if you do some variation of:

class C
{
  enum x = __traits(xxSymbol, C).length;
}

There was a series of cases like this which determineVtbl solved, but I'm not sure they add value, and a forward-reference error is probably fine?