[lldb] Change the type alias resolution algorithm #10036
Conversation
adrian-prantl
requested review from
augusto2112 and
kastiglione
and removed request for
a team
adrian-prantl
force-pushed
the
144884987
branch
from
227dceb to
d220d9d
Compare
|
@swift-ci test |
adrian-prantl
force-pushed
the
144884987
branch
from
d220d9d to
8d29597
Compare
| default: | ||
| break; | ||
| default: { | ||
| llvm::SmallVector<NodePointer, 2> children; |
There was a problem hiding this comment.
Is this code meant to both be here and in GetCanonicalNode?
| @@ -514,12 +522,6 @@ class TypeSystemSwiftTypeRef : public TypeSystemSwift { | |||
| CompilerType LookupClangForwardType(llvm::StringRef name, | |||
| llvm::ArrayRef<CompilerContext> decl_context); | |||
|
|
|||
| /// Recursively resolves all type aliases. | |||
| swift::Demangle::NodePointer | |||
There was a problem hiding this comment.
ResolveAllTypeAliases is gone from the header but not implementation?
| return TypeSystemSwiftTypeRef::Transform(dem, node, [&](NodePointer node) { | ||
| return Canonicalize(dem, node, flavor); | ||
| }); | ||
| NodePointer transformed = Canonicalize(dem, node, flavor); |
There was a problem hiding this comment.
Should this block of code be in a separate function called TransformPreOrder?
There was a problem hiding this comment.
There is only a single use, and it has shortcut in that it won't traverse the children of a transformed node, which is something a generic preorder transform maybe shouldn't do, so I decided to punt on that until we have a second use-case.
| /// Desugar to this node and if it is a type alias resolve it by | ||
| /// looking up its type in the debug info. | ||
| swift::Demangle::NodePointer | ||
| Canonicalize(swift::Demangle::Demangler &dem, |
There was a problem hiding this comment.
Why were these functions moved to be public? I don't see you using them anywhere outside TypeSystemSwiftTypeRef
adrian-prantl
force-pushed
the
144884987
branch
from
8d29597 to
edfc421
Compare
from a post-oder to a pre-order traversal. This is necessary to resolve generic type aliases that bind other type aliases in one go, instead of first resolving the bound type aliases. Debug Info will have a record for `SomeAlias<SomeOtherAlias>` but not `SomeAlias<WhatSomeOtherAliasResolvesTo>` because it tries to preserve all sugar. rdar://144884987
adrian-prantl
force-pushed
the
144884987
branch
from
edfc421 to
b16a454
Compare
|
@swift-ci test |
|
@swift-ci test |
from a post-oder to a pre-order traversal. This is necessary to resolve generic type aliases that bind other type aliases in one go, instead of first resolving the bound type aliases. Debug Info will have a record for
SomeAlias<SomeOtherAlias>but notSomeAlias<WhatSomeOtherAliasResolvesTo>because it tries to preserve all sugar.rdar://144884987