Skip to content

[Sema] NFC: fix assert-only crasher with overloaded generics #22689

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Feb 19, 2019
Merged

[Sema] NFC: fix assert-only crasher with overloaded generics #22689

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 3 additions & 1 deletion lib/Sema/CSDiagnostics.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -136,7 +136,9 @@ ValueDecl *RequirementFailure::getDeclRef() const {
auto *anchor = getRawAnchor();
auto *locator = cs.getConstraintLocator(anchor);
if (auto *AE = dyn_cast<CallExpr>(anchor)) {
assert(isa<TypeExpr>(AE->getFn()));
// NOTE: In valid code, the function can only be a TypeExpr
assert(isa<TypeExpr>(AE->getFn()) ||
isa<OverloadedDeclRefExpr>(AE->getFn()));
ConstraintLocatorBuilder ctor(locator);
locator = cs.getConstraintLocator(
ctor.withPathElement(PathEltKind::ApplyFunction)
Expand Down
13 changes: 13 additions & 0 deletions validation-test/Sema/type_checker_crashers_fixed/overloaded_generic.swift
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,13 @@
// RUN: not %target-swift-frontend %s -typecheck

protocol A {}
protocol B {}

struct X : A {}
struct Y : B {}

struct G<T:A> { var v : T }
struct G<T:B> { var v : T }

let a = G(v: X())
let b = G(v: Y())
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Updated assert makes sense, thank you! But the example itself doesn't trigger updated code in requirement failure, or any diagnostics for that matter. Would be great to get this example to diagnose before merging. I think that might be a bit tough though because fixing requirements in overloaded decls like that usually leads to ambiguities because both Gs would much.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hi @xedin – If you look at the test case, it negates the exit value via not because the actual diagnostics doesn't matter, just that the test case should no longer crash the compiler.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sorry what I mean is example as-is type-checks correctly, it's not going to generate any fixes, and subsequently not going to trigger getDeclRef() code and crash.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The test case does not type check as is. Before this pull request:

/tmp/gt.swift:8:8: error: invalid redeclaration of 'G'
struct G<T:B> { var v : T }
       ^
/tmp/gt.swift:7:8: note: 'G' previously declared here
struct G<T:A> { var v : T }
       ^
/tmp/gt.swift:11:9: error: generic struct 'G' requires that 'Y' conform to 'A'
let b = G(v: Y())
        ^
/tmp/gt.swift:7:8: note: where 'T' = 'Y'
struct G<T:A> { var v : T }
       ^
swift: /home/dave/s/u/swift/lib/Sema/CSDiagnostics.cpp:149: swift::ValueDecl *swift::constraints::RequirementFailure::getDeclRef() const: Assertion `isa<TypeExpr>(AE->getFn())' failed.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ah, sorry for the confusion, I've tried in REPL and it works differently there, would create two decls in different modules, that's why it type-checked just fine...