[tests] Compilation server vs defineModule/defineType

[kLabz]

Goal

• Find a proper solution for a compilation server safe way of adding types with defineModule/defineType
• Have a test in Haxe repo to make sure we don't break that with compilation server changes
• Add a cookbook entry showing how to implement it
• Point library authors to that cookbook entry if they have "x is redefined from x" / missing types in generated code issues

Provided implementation

Feel free to suggest changes to the implementation I provided, as I plan to put in on the cookbook.
I did not manage to break it, but if you find a way please tell me and I'll update the tests and try to address the issue.

I also looked into implementing CompilationServer.isAlive(t:Type):Bool but it wasn't working as well as the custom implementation I'm using here in the end...

Note: name mangling is very minimal there and won't be enough in some (many) cases; will add something about it in cookbook to tell people to use something that will generate truly unique names for their use case (any suggestion on a particular haxe.macro.Type name mangling implementation?)

[kLabz]

... I think current failures might actually be false negatives.

I managed to reproduce a few failures locally, and this is the debug data:

src/cases/CsSafeTypeBuilding.hx:45: Fail: doesn't contain "[runtime] Hello from Foo__Bar__Bar"
src/cases/CsSafeTypeBuilding.hx:46: Fail: doesn't contain "[runtime] Hello from Foo__Baz__Baz"
src/cases/CsSafeTypeBuilding.hx:51: Haxe print: Building Baz.
src/cases/CsSafeTypeBuilding.hx:51: Haxe print: Building Main.
src/cases/CsSafeTypeBuilding.hx:51: Haxe print: Building Bar.
src/cases/CsSafeTypeBuilding.hx:51: Haxe print: [Bar] Generating type for Foo__Bar__Bar.
src/cases/CsSafeTypeBuilding.hx:51: Haxe print: [Main] Generating type for Foo__Main__Main.
src/cases/CsSafeTypeBuilding.hx:51: Haxe print: [Baz] Generating type for Foo__Baz__Baz.
src/cases/CsSafeTypeBuilding.hx:51: Haxe print: [Baz] Reusing previously generated type for Foo__Bar__Bar.
src/cases/CsSafeTypeBuilding.hx:51: Haxe print: [runtime] Hello from Bar
src/cases/CsSafeTypeBuilding.hx:51: Haxe print: 
src/cases/CsSafeTypeBuilding.hx:51: Haxe print: [runtime] Hello from Baz
src/cases/CsSafeTypeBuilding.hx:51: Haxe print: 
src/cases/CsSafeTypeBuilding.hx:51: Haxe print: [runtime] Hello from Foo__Bar__Bar[runtime] Hello from Foo__Baz__Baz
src/cases/CsSafeTypeBuilding.hx:51: Haxe print: [runtime] Hello from Foo__Main__Main
src/cases/CsSafeTypeBuilding.hx:51: Haxe print: 
src/cases/CsSafeTypeBuilding.hx:51: Haxe print: [runtime] Hello from Main
src/cases/CsSafeTypeBuilding.hx:51: Haxe print: 

The prints that were not found were actually there, but some have been merged into a single "message"...
Checking if CI failures are the same.

[Simn]

Mac-specific JS failures? Huh...

[kLabz]

Nope it's not mac specific failures, can happen on linux too.

[kLabz]

OK, I got it into control

Haxe server prints are passed as 0x01 + message + newline
haxeserver parses prints as such, but sometimes we get 0x01 + message1 + 0x01 + message2 + newline

Fixed in #11459

[kLabz]

Works on 4.3.x too but requires #11153 to be added to 4.3_bugfix