Check access control for the generic requirements of subscripts and typealiases

[jrose-apple]

Oops.

This is a little more involved than it should have been because the requirements on a generic typealias don't always carry a Type anymore; sometimes all you have is the TypeRepr. That should still be okay in practice as long as we don't start doing that for var/let, which can have part of a type be inferred but not all of it.

We don't have an idiom for "this should be an error in the next -swift-version" yet. Maybe it should be 9999, like we're using in the standard library for Apple OS availability?

[jrose-apple]

@swift-ci Please test

[jrose-apple]

@swift-ci Please test source compatibility

[jrose-apple]

@swift-ci Please smoke test compiler performance

[slavapestov]

I think long term, we'll want this kind of thing to kick off requests to get referenced decls from where clauses, etc. Right now there's an ordering dependency between having evaluated certain requests and running the code here because it relies in the decl being bound in the TypeReprs, which is fragile. Ideally we'd remove the bound decl from IdentTypeRepr altogether.

[jrose-apple]

The where clause logic is already doing that, but the inherited types aren't going through that logic.

[swift-ci]

Summary for master smoketest

Unexpected test results, excluded stats for ReactiveCocoa

No regressions above thresholds

• Debug
  • debug brief
  • debug detailed
• Release
  • release brief
  • release detailed

Debug

debug brief

Regressed (0)

name	old	new	delta	delta_pct

Improved (0)

name	old	new	delta	delta_pct

Unchanged (delta < 1.0% or delta < 100.0ms) (3)

name	old	new	delta	delta_pct
Frontend.NumInstructionsExecuted	99,830,338,602	99,823,975,873	-6,362,729	-0.01%
LLVM.NumLLVMBytesOutput	6,152,964	6,152,916	-48	-0.0%
time.swift-driver.wall	13.0s	12.9s	-66.0ms	-0.51%

debug detailed

Regressed (0)

name	old	new	delta	delta_pct

Improved (0)

name	old	new	delta	delta_pct

Unchanged (delta < 1.0% or delta < 100.0ms) (23)

name	old	new	delta	delta_pct
AST.NumImportedExternalDefinitions	1,032	1,032	0	0.0%
AST.NumLoadedModules	828	828	0	0.0%
AST.NumTotalClangImportedEntities	4,210	4,210	0	0.0%
AST.NumUsedConformances	886	886	0	0.0%
IRModule.NumIRBasicBlocks	18,032	18,032	0	0.0%
IRModule.NumIRFunctions	10,542	10,542	0	0.0%
IRModule.NumIRGlobals	8,013	8,013	0	0.0%
IRModule.NumIRInsts	313,099	313,099	0	0.0%
IRModule.NumIRValueSymbols	17,618	17,618	0	0.0%
LLVM.NumLLVMBytesOutput	6,152,964	6,152,916	-48	-0.0%
SILModule.NumSILGenFunctions	5,370	5,370	0	0.0%
SILModule.NumSILOptFunctions	7,041	7,041	0	0.0%
Sema.NumConformancesDeserialized	16,171	16,171	0	0.0%
Sema.NumConstraintScopes	40,312	40,312	0	0.0%
Sema.NumDeclsDeserialized	134,587	134,587	0	0.0%
Sema.NumDeclsValidated	11,208	11,208	0	0.0%
Sema.NumFunctionsTypechecked	2,516	2,516	0	0.0%
Sema.NumGenericSignatureBuilders	4,776	4,776	0	0.0%
Sema.NumLazyGenericEnvironments	30,003	30,003	0	0.0%
Sema.NumLazyGenericEnvironmentsLoaded	2,012	2,012	0	0.0%
Sema.NumLazyIterableDeclContexts	23,862	23,862	0	0.0%
Sema.NumTypesDeserialized	59,934	59,934	0	0.0%
Sema.NumTypesValidated	11,902	11,902	0	0.0%

Release

release brief

Regressed (0)

name	old	new	delta	delta_pct

Improved (0)

name	old	new	delta	delta_pct

Unchanged (delta < 1.0% or delta < 100.0ms) (3)

name	old	new	delta	delta_pct
Frontend.NumInstructionsExecuted	142,639,098,985	142,570,846,346	-68,252,639	-0.05%
LLVM.NumLLVMBytesOutput	7,168,728	7,168,716	-12	-0.0%
time.swift-driver.wall	23.5s	23.5s	-27.6ms	-0.12%

release detailed

Regressed (0)

name	old	new	delta	delta_pct

Improved (0)

name	old	new	delta	delta_pct

Unchanged (delta < 1.0% or delta < 100.0ms) (23)

name	old	new	delta	delta_pct
AST.NumImportedExternalDefinitions	402	402	0	0.0%
AST.NumLoadedModules	76	76	0	0.0%
AST.NumTotalClangImportedEntities	2,146	2,146	0	0.0%
AST.NumUsedConformances	890	890	0	0.0%
IRModule.NumIRBasicBlocks	21,170	21,170	0	0.0%
IRModule.NumIRFunctions	10,034	10,034	0	0.0%
IRModule.NumIRGlobals	8,107	8,107	0	0.0%
IRModule.NumIRInsts	224,230	224,230	0	0.0%
IRModule.NumIRValueSymbols	17,289	17,289	0	0.0%
LLVM.NumLLVMBytesOutput	7,168,728	7,168,716	-12	-0.0%
SILModule.NumSILGenFunctions	4,176	4,176	0	0.0%
SILModule.NumSILOptFunctions	5,849	5,849	0	0.0%
Sema.NumConformancesDeserialized	12,457	12,457	0	0.0%
Sema.NumConstraintScopes	39,382	39,382	0	0.0%
Sema.NumDeclsDeserialized	32,445	32,445	0	0.0%
Sema.NumDeclsValidated	7,874	7,874	0	0.0%
Sema.NumFunctionsTypechecked	2,204	2,204	0	0.0%
Sema.NumGenericSignatureBuilders	1,718	1,718	0	0.0%
Sema.NumLazyGenericEnvironments	6,871	6,871	0	0.0%
Sema.NumLazyGenericEnvironmentsLoaded	130	130	0	0.0%
Sema.NumLazyIterableDeclContexts	4,259	4,259	0	0.0%
Sema.NumTypesDeserialized	18,144	18,144	0	0.0%
Sema.NumTypesValidated	6,942	6,942	0	0.0%

[beccadax]

LGTM.

[beccadax]

I wouldn't do it in this PR even if you hadn't merged it already, but it might make sense to build something in to the DiagnosticEngine that downgrades errors to warnings if the language version is not high enough.

TC.diagnoseLanguageChange(/*after:*/5, owner, diag::generic_param_usable_from_inline ...)

[jrose-apple]

Yeah, that's come up before. The main barrier in this case is that the text for these is slightly different (one is phrased as "oh, you shouldn't do this" and the other is "this doesn't work"), but maybe that could be abstracted out too.