Trailing commas

[RblSb]

Second commit fixes all tests, except for one with Unexpected } error instead of new Expected , or ) output for trace(0, line. Would be nice to see some Simon's dark magic to make call(a,) possible too...

[kLabz]

I thought we didn't want that 🤔 Maybe I'm misremembering...

[kLabz]

I seem to remember incorrectly. However, why this PR instead of #10159 ?

[RblSb]

Yea, i just rebased that PR with fixed conficts, and found part that breaks diagnostics tests

[kLabz]

@Aurel300 any comment about that change maybe?

https://github.com/HaxeFoundation/haxe/pull/11009/files#diff-c9e4d364c858fe7652c71366fa690f05f38e6d6ad6c2e0e67a0e4f2fea1a0afbL1672-R1683

[Aurel300]

Thanks for taking over the PR!

Regarding the one failing test: I honestly don't see why the expected error message is any better than the new error message.

I don't remember what the falling test cases were showing with my PR (and the logs are now gone): were they also just about different error messages? Trailing commas in calls seem extremely useful to me, I wouldn't want to omit these.

Regarding the change @kLabz linked: it would be better to solve the test case failures of course, but otherwise I see no point in keeping the implementation commented out like this.

[RblSb]

If you uncomment that block, there is many signature errors with no completion point, like this, full logs:
https://github.com/RblSb/haxe/actions/runs/4366316441/jobs/7636448711

And no Enum constructors in argument completion (with fallback to basic toplevel in vshaxe):

haxe/tests/display/src/cases/Toplevel.hx

Lines 226 to 236 in 3d52330

	/**
	class Main {
	static function f(t:Type.ValueType) {}
	public static function main() {
	f({-1-}
	**/
	function testExpectedType1() {
	var fields = toplevel(pos(1));
	eq(true, hasToplevel(fields, "enum", "TNull"));
	}

Have no idea how added Comma > PClose case can change display logic, since there is no Comma in f(| to match at all... f(|) still collects enums here btw.

[RblSb]

Feeling lucky today. @Simn what do you think?

[kLabz]

Are cases of (,) caught? I feel they should not be allowed

[RblSb]

Thanks!

[Aurel300]

Note that the final task of the original PR still remains: the TextMate syntax needs to be updated to support this, otherwise syntax highlighting (in the IDE, here on GitHub, etc) gets broken after a trailing comma. I remember this being rather annoying due to how the current syntax highlighting is structured.