Change the underlying type of TypeAliasDecls to an interface type #6321

slavapestov · 2016-12-16T07:03:25Z

No description provided.

…ying type - TypeAliasDecl::getAliasType() is gone. Now, getDeclaredInterfaceType() always returns the NameAliasType. - NameAliasTypes now always desugar to the underlying type as an interface type. - The NameAliasType of a generic type alias no longer desugars to an UnboundGenericType; call TypeAliasDecl::getUnboundGenericType() if you want that. - The "lazy mapTypeOutOfContext()" hack for deserialized TypeAliasDecls is gone. - The process of constructing a synthesized TypeAliasDecl is much simpler now; instead of calling computeType(), setInterfaceType() and then setting the recursive properties in the right order, just call setUnderlyingType(), passing it either an interface type or a contextual type. In particular, many places weren't setting the recursive properties, such as the ClangImporter and deserialization. This meant that queries such as hasArchetype() or hasTypeParameter() would return incorrect results on NameAliasTypes, which caused various subtle problems. - Finally, add some more tests for generic typealiases, most of which fail because they're still pretty broken.

slavapestov · 2016-12-16T07:03:41Z

@swift-ci Please test

slavapestov · 2016-12-16T07:05:57Z

lib/Serialization/Deserialization.cpp

    declOrOffset = alias;

+    // FIXME: Do we need to read a GenericEnvironment even if we don't
+    // have our own generic parameters? The typealias might itself be in
+    // generic context.


@DougGregor This looks like a bug, but changing it breaks things...

slavapestov · 2016-12-16T07:11:01Z

test/api-digester/source-stability.swift.expected

 Var Dictionary.endIndex has declared type change from DictionaryIndex<Key, Value> to Dictionary<Key, Value>.Index
 Var Dictionary.startIndex has declared type change from DictionaryIndex<Key, Value> to Dictionary<Key, Value>.Index
 Var Set.endIndex has declared type change from SetIndex<Element> to Set<Element>.Index
 Var Set.startIndex has declared type change from SetIndex<Element> to Set<Element>.Index
+Constructor BidirectionalSlice.init(base:bounds:) has 2nd parameter type change from Range<Base.Index> to Range<BidirectionalSlice.Index>


@nkcsgexi I'm really sorry for constantly breaking the output of this tool, but we need to figure out a way to canonicalize types better.

In particular, it seems we need a better strategy for handling changes where the serialized type resolves to an alias of the same type as before, or a generic parameter that is related to the previous one by a same-type constraint. I'm at a loss as to how to work this into the model of "dump API of two serialized module files, diff them". We might need a more complex approach.

Or, since with this patch the interface type changeover is mostly complete, we can start again with a fresh dump for the "known good" API.

It is surely ok to break the output, but would you please document that they are false positives? As an alternative, we should have a universal clang importer (and deserializer) in the future that can load any versions of modules and compare them in a fuller context. But, as for now, as you said, it seems too hard to eliminate false positives like these.

swift-ci · 2016-12-16T07:13:54Z

Build failed
Jenkins build - Swift Test Linux Platform
Git Commit - 2c6b9f7
Test requested by - @slavapestov

slavapestov · 2016-12-16T07:20:42Z

Please test with following PR:
apple/swift-lldb#107

@swift-ci Please test Linux

swift-ci · 2016-12-16T07:23:55Z

Build failed
Jenkins build - Swift Test Linux Platform
Git Commit - 2c6b9f7
Test requested by - @slavapestov

slavapestov · 2016-12-16T07:26:04Z

Please test with following PR:
apple/swift-lldb#107

@swift-ci Please test Linux

swift-ci · 2016-12-16T07:32:52Z

Build failed
Jenkins build - Swift Test OS X Platform
Git Commit - 2c6b9f7
Test requested by - @slavapestov

swift-ci · 2016-12-16T07:39:21Z

Build failed
Jenkins build - Swift Test Linux Platform
Git Commit - 2c6b9f7
Test requested by - @slavapestov

slavapestov · 2016-12-16T07:44:04Z

Please test with following PR:
apple/swift-lldb#107

@swift-ci Please test Linux

slavapestov · 2016-12-16T07:53:23Z

Hmm...

23:29:18 0  swift                    0x000000011149ea06 SignalHandler(int) + 454
23:29:18 1  libsystem_platform.dylib 0x00007fff84ba652a _sigtramp + 26
23:29:18 2  swift                    0x000000010ecd2d08 swift::ArchetypeBuilder::resolveArchetype(swift::Type) + 72
23:29:18 3  swift                    0x000000010ecd2d78 swift::ArchetypeBuilder::resolveArchetype(swift::Type) + 184
23:29:18 4  swift                    0x000000010ecd2d78 swift::ArchetypeBuilder::resolveArchetype(swift::Type) + 184
23:29:18 5  swift                    0x000000010ecd2d78 swift::ArchetypeBuilder::resolveArchetype(swift::Type) + 184
23:29:18 6  swift                    0x000000010ecd4de3 swift::ArchetypeType::resolveNestedType(std::__1::pair<swift::Identifier, swift::Type>&) const + 51
23:29:18 7  swift                    0x000000010ede78e2 swift::ArchetypeType::getAllNestedTypes(bool) const + 66
23:29:18 8  swift                    0x000000010e4cd3b8 profileArchetypeConstraints(swift::Type, llvm::FoldingSetNodeID&, llvm::DenseMap<swift::ArchetypeType*, unsigned int, llvm::DenseMapInfo<swift::ArchetypeType*>, llvm::detail::DenseMapPair<swift::ArchetypeType*, unsigned int> >&) + 552
23:29:18 9  swift                    0x000000010e4cd3de profileArchetypeConstraints(swift::Type, llvm::FoldingSetNodeID&, llvm::DenseMap<swift::ArchetypeType*, unsigned int, llvm::DenseMapInfo<swift::ArchetypeType*>, llvm::detail::DenseMapPair<swift::ArchetypeType*, unsigned int> >&) + 590
23:29:18 10 swift                    0x000000010e4cd3de profileArchetypeConstraints(swift::Type, llvm::FoldingSetNodeID&, llvm::DenseMap<swift::ArchetypeType*, unsigned int, llvm::DenseMapInfo<swift::ArchetypeType*>, llvm::detail::DenseMapPair<swift::ArchetypeType*, unsigned int> >&) + 590
23:29:18 11 swift                    0x000000010e4cd4d2 swift::irgen::TypeConverter::getExemplarArchetype(swift::ArchetypeType*) + 114
23:29:18 12 swift                    0x000000010e4d4687 swift::Type llvm::function_ref<swift::Type (swift::Type)>::callback_fn<swift::irgen::TypeConverter::getExemplarType(swift::CanType)::$_0>(long, swift::Type) + 23
23:29:18 13 swift                    0x000000010ede91c4 swift::Type::transform(llvm::function_ref<swift::Type (swift::Type)>) const + 68
23:29:18 14 swift                    0x000000010ede94c7 swift::Type::transform(llvm::function_ref<swift::Type (swift::Type)>) const + 839
23:29:18 15 swift                    0x000000010e4cd62d swift::irgen::TypeConverter::getExemplarType(swift::CanType) + 93

slavapestov · 2016-12-16T07:58:20Z

@swift-ci Please test macOS

swift-ci · 2016-12-16T08:01:42Z

Build failed
Jenkins build - Swift Test Linux Platform
Git Commit - 2c6b9f7
Test requested by - @slavapestov

slavapestov · 2016-12-16T08:06:22Z

Please test with following PR:
apple/swift-lldb#107

@swift-ci Please test Linux

swift-ci · 2016-12-16T08:19:27Z

Build failed
Jenkins build - Swift Test Linux Platform
Git Commit - 2c6b9f7
Test requested by - @slavapestov

slavapestov · 2016-12-16T09:11:51Z

Please test with following PR:
apple/swift-lldb#107

@swift-ci Please test Linux

slavapestov added 4 commits December 15, 2016 22:46

Serialization: Add an assertion

ae3d282

Don't call VarDecl::getType() on deserialized VarDecls

a384b2a

AST: Remove the hack for lazily setting contextual types on VarDecls

757f253

slavapestov commented Dec 16, 2016

View reviewed changes

slavapestov merged commit 2ff9061 into swiftlang:master Dec 16, 2016

ahoppen mentioned this pull request Jan 25, 2023

Improve diagnostics for effect specifiers swiftlang/swift-syntax#1260

Merged

Change the underlying type of TypeAliasDecls to an interface type #6321

Change the underlying type of TypeAliasDecls to an interface type #6321

Uh oh!

Conversation

slavapestov commented Dec 16, 2016

Uh oh!

slavapestov commented Dec 16, 2016

Uh oh!

slavapestov Dec 16, 2016

Choose a reason for hiding this comment

Uh oh!

slavapestov Dec 16, 2016

Choose a reason for hiding this comment

Uh oh!

nkcsgexi Dec 16, 2016

Choose a reason for hiding this comment

Uh oh!

swift-ci commented Dec 16, 2016

Uh oh!

slavapestov commented Dec 16, 2016

Uh oh!

swift-ci commented Dec 16, 2016

Uh oh!

slavapestov commented Dec 16, 2016

Uh oh!

swift-ci commented Dec 16, 2016

Uh oh!

swift-ci commented Dec 16, 2016

Uh oh!

slavapestov commented Dec 16, 2016

Uh oh!

slavapestov commented Dec 16, 2016

Uh oh!

slavapestov commented Dec 16, 2016

Uh oh!

swift-ci commented Dec 16, 2016

Uh oh!

slavapestov commented Dec 16, 2016

Uh oh!

swift-ci commented Dec 16, 2016

Uh oh!

slavapestov commented Dec 16, 2016

Uh oh!

Uh oh!