Simon Krajewski
Simon Krajewski
I have checked hxcpp again and both tests work there now. I also made sure that the analyzer doesn't inadvertently fix it. I didn't test HL because all I ever...
[My CI](https://travis-ci.org/Simn/haxe/jobs/374343098) says this fails on C# and PHP of all things.
I confirmed HL to still be failing here as well.
@RealyUniqueName Could you look at the PHP failure? ```haxe class Main { @:analyzer(ignore) static public function main() { var a:Array = [1, 2, 3]; var b:Array = cast a; b.push(4);...
What problems does `erase_generics` cause?
I've updated the branch again: * **PHP**: `src/unit/TestArray.hx:12: 4 should be null` * **HL**: `src/unit/TestArray.hx:9: characters 5-11 : Don't know how to compare #hl.types.ArrayDyn and #hl.types.ArrayBytes_Int` * **C#**: ```src/unit/TestArray.hx:9: false...
Sure, I keep these branches on my own GitHub because that tends to be better for CI: https://github.com/Simn/haxe/tree/array_variance_tests
Yes but is that for me or for you to fix? If it's a problem with temp vars we have to check if it can be reproduced manually (without inline).
Then we have to mark the affected types with some metadata. Though I don't really like having to support copy-semantics just for this. Note that this example also doesn't behave...
This would mean that for certain types we never create a temp var. This is quite problematic for various cases I can think of. IMO this has to be approached...