[ast]/[silgen] If a case stmt is a fallthrough source, tail allocate a pointer in the case stmt to the fallthrough case. #23284
Conversation
…il allocated pointer field. rdar://47467128
|
I am mainly asking for review of the first commit. |
|
@swift-ci smoke test |
| @@ -920,6 +922,45 @@ class CaseLabelItem { | |||
| } | |||
| }; | |||
|
|
|||
| /// FallthroughStmt - The keyword "fallthrough". | |||
| class FallthroughStmt : public Stmt { | |||
There was a problem hiding this comment.
There are no changes here to FallthroughStmt. I just ran into issues with my uses for it in CaseStmt and then discovered that I couldn't just use a forward declaration of it = /.
There was a problem hiding this comment.
*So I moved it above CaseStmt.
|
Since I based the logic on identifying fallthroughs on the logic in SILGenPattern, the actual output from SILGenPattern should be not changed at all so test updates were not needed. |
| private llvm::TrailingObjects<CaseStmt, CaseLabelItem> { | ||
| class CaseStmt final | ||
| : public Stmt, | ||
| private llvm::TrailingObjects<CaseStmt, FallthroughStmt *, |
There was a problem hiding this comment.
I decided to store the entire FallthroughStmt * rather than a pointer to the CaseStmt* returned from FallthroughStmt::getFallthroughDest(). This seems more flexible. That being said, I did not want to add it to the API surface, so I did not expose it directly.
| @@ -943,24 +986,47 @@ class CaseStmt final : public Stmt, | |||
|
|
|||
| llvm::PointerIntPair<Stmt *, 1, bool> BodyAndHasBoundDecls; | |||
|
|
|||
| /// Set to true if we have a fallthrough. | |||
| /// | |||
| /// TODO: Once we have CaseBodyVarDecls, use the bit in BodyAndHasBoundDecls | |||
There was a problem hiding this comment.
To be clear here: HasBoundDecls is true iff we have CaseBodyVarDecls to my commit adding those will fix that. I just did not want to create a field called BodyAndHasBoundDeclsAndHasFallthrough...
|
One last comment. I checked for other places that were using CaseStmt::create. This is the only call site AFAIKT that can have a fallthrough. This can be seen since FallthroughStmt are only created in the parser, never in sema. |
…est() instead of computing this itself. rdar://47467128
gottesmm
force-pushed
the
pr-0b4de614fe103f8e5a161be5219ba2cb7148aa25
branch
from
91c34cc to
7b27b4b
Compare
|
@swift-ci smoke test |
1 similar comment
|
@swift-ci smoke test |
|
@swift-ci smoke test and merge |
|
I killed the two non smoke test and merge jobs that I triggered. |
|
The linux case failed b/c of 12:19:01 SwiftXCTestFunctionalTests :: Asynchronous/Expectations/main.swift I need to get around to fixing that. |
|
@swift-ci smoke test linux platform |
The second commit uses this in SILGenPattern in a trivial way to show that everything works.