Optimize attribute checking

[kerams]

TcTypeAndRecover would redo ResolveTypeLongIdent, so we bypass duplicate work and skip directly to TcTypeApp. This has the additional benefit that we no longer capture 2 Item.Types name resolutions for each attribute (one with ItemOccurence.UseInAttribute and the other ItemOccurence.UseInType) but only the first one.

[kerams]

I also wonder whether CheckTyconAccessible cenv.amap m env.AccessRights tcref in TcTypeApp is strictly necessary. tcref will almost certainly have come from ResolveTypeLongIdent or similar, which should perform accessibility checks. Not sure how safe it would be to get rid of this.

[smoothdeveloper]

@kerams, wondering if we could also optimize which lookup to try first, if the text already ends with Attribute, 99.99% of cases it would be better to try the one without appending it.

I can only see weird stuff like AttributeAttribute that would require the secondary lookup then.

[kerams]

That would give a different meaning to the program if you have SomeAttributeAttribute and SomeAttribute in scope.

[smoothdeveloper]

@kerams, ok makes sense and it is a bit sad edge case :(

I'm wondering if TcTypeMeasureApp is equivalent to your TcTypeApp? does it still type check measures annotated values passed to an attribute?

openFSharp.Data.UnitSystems.SI.UnitNamestypeAttAttribute(a: int<second>)=inherit System.Attribute()[<Att(1)>]typeFoo()=memberx.X=1

[vzarytovskii]

This will have to wait unfortunately, net8 is now only open for fixes.

[vzarytovskii]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 2 pipeline(s).