[Clang][Sema] Expose static inline functions from GMF #104701
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
They should be exposed even when two-phase name lookup and ADL are involved.
llvmbot
added
clang
|
@llvm/pr-subscribers-clang @llvm/pr-subscribers-clang-modules Author: Jan Kokemüller (jiixyj) ChangesIn C, it is a common pattern to have Similarly, in C++, sometimes people use templates with internal linkage in headers. As the GMF was designed to be a transitional mechanism for headers, special case those functions in This fixes <#98021>. Full diff: https://github.com/llvm/llvm-project/pull/104701.diff 6 Files Affected:
diff --git a/clang/lib/Sema/SemaOverload.cpp b/clang/lib/Sema/SemaOverload.cpp
index 52f640eb96b73b..ca070d0e22d472 100644
--- a/clang/lib/Sema/SemaOverload.cpp
+++ b/clang/lib/Sema/SemaOverload.cpp
@@ -6926,11 +6926,26 @@ void Sema::AddOverloadCandidate(
/// have linkage. So that all entities of the same should share one
/// linkage. But in clang, different entities of the same could have
/// different linkage.
- NamedDecl *ND = Function;
- if (auto *SpecInfo = Function->getTemplateSpecializationInfo())
+ const NamedDecl *ND = Function;
+ bool IsImplicitlyInstantiated = false;
+ if (auto *SpecInfo = Function->getTemplateSpecializationInfo()) {
ND = SpecInfo->getTemplate();
-
- if (ND->getFormalLinkage() == Linkage::Internal) {
+ IsImplicitlyInstantiated = SpecInfo->getTemplateSpecializationKind() ==
+ TSK_ImplicitInstantiation;
+ }
+
+ /// Don't remove inline functions with internal linkage from the overload
+ /// set if they are declared in a GMF.
+ /// The global module is meant to be a transition mechanism for C and C++
+ /// headers.
+ /// Inline functions with internal linkage are a common pattern in headers
+ /// to avoid ODR issues.
+ const bool IsInlineFunctionInGMF =
+ Function->getOwningModule() &&
+ Function->getOwningModule()->isGlobalModule() &&
+ (IsImplicitlyInstantiated || Function->isInlined());
+
+ if (ND->getFormalLinkage() == Linkage::Internal && !IsInlineFunctionInGMF) {
Candidate.Viable = false;
Candidate.FailureKind = ovl_fail_module_mismatched;
return;
diff --git a/clang/test/Modules/expose-static-inline-from-gmf-1.cppm b/clang/test/Modules/expose-static-inline-from-gmf-1.cppm
new file mode 100644
index 00000000000000..4de9b583dac8da
--- /dev/null
+++ b/clang/test/Modules/expose-static-inline-from-gmf-1.cppm
@@ -0,0 +1,37 @@
+// RUN: rm -rf %t
+// RUN: mkdir -p %t
+// RUN: split-file %s %t
+//
+// RUN: %clang -std=c++20 %t/a.cppm --precompile -o %t/a.pcm \
+// RUN: -DTEST_INLINE
+// RUN: %clang -std=c++20 %t/test.cc -fprebuilt-module-path=%t -fsyntax-only -Xclang -verify \
+// RUN: -DTEST_INLINE
+//
+// RUN: %clang -std=c++20 %t/a.cppm --precompile -o %t/a.pcm
+// RUN: %clang -std=c++20 %t/test.cc -fprebuilt-module-path=%t -fsyntax-only -Xclang -verify
+
+//--- a.h
+#ifdef TEST_INLINE
+#define INLINE inline
+#else
+#define INLINE
+#endif
+static INLINE void func(long) {}
+template <typename T = long> void a() { func(T{}); }
+
+//--- a.cppm
+module;
+#include "a.h"
+export module a;
+export using ::a;
+
+//--- test.cc
+import a;
+auto m = (a(), 0);
+
+#ifdef TEST_INLINE
+// expected-no-diagnostics
+#else
+// [email protected]:7 {{no matching function for call to 'func'}}
+// [email protected]:2 {{in instantiation of function template specialization 'a<long>' requested here}}
+#endif
diff --git a/clang/test/Modules/expose-static-inline-from-gmf-2.cppm b/clang/test/Modules/expose-static-inline-from-gmf-2.cppm
new file mode 100644
index 00000000000000..c89b613f5074b1
--- /dev/null
+++ b/clang/test/Modules/expose-static-inline-from-gmf-2.cppm
@@ -0,0 +1,22 @@
+// RUN: rm -rf %t
+// RUN: mkdir -p %t
+// RUN: split-file %s %t
+//
+// RUN: %clang -std=c++20 %t/a.cppm --precompile -o %t/a.pcm
+// RUN: %clang -std=c++20 %t/test.cc -fprebuilt-module-path=%t -fsyntax-only -Xclang -verify
+
+//--- a.h
+template <typename G> static inline void func() {}
+template <typename T = long> void a() { func<T>(); }
+
+//--- a.cppm
+module;
+#include "a.h"
+export module a;
+export using ::a;
+
+//--- test.cc
+import a;
+auto m = (a(), 0);
+
+// expected-no-diagnostics
diff --git a/clang/test/Modules/expose-static-inline-from-gmf-3.cppm b/clang/test/Modules/expose-static-inline-from-gmf-3.cppm
new file mode 100644
index 00000000000000..dee7cddafdf701
--- /dev/null
+++ b/clang/test/Modules/expose-static-inline-from-gmf-3.cppm
@@ -0,0 +1,24 @@
+// RUN: rm -rf %t
+// RUN: mkdir -p %t
+// RUN: split-file %s %t
+//
+// RUN: %clang -std=c++20 %t/a.cppm --precompile -o %t/a.pcm
+// RUN: %clang -std=c++20 %t/test.cc -fprebuilt-module-path=%t -fsyntax-only -Xclang -verify
+
+//--- a.h
+namespace ns {
+template <typename G> static void func() {}
+template <typename T = long> void a() { func<T>(); }
+}
+
+//--- a.cppm
+module;
+#include "a.h"
+export module a;
+export using ns::a;
+
+//--- test.cc
+import a;
+auto m = (a(), 0);
+
+// expected-no-diagnostics
diff --git a/clang/test/Modules/expose-static-inline-from-gmf-4.cppm b/clang/test/Modules/expose-static-inline-from-gmf-4.cppm
new file mode 100644
index 00000000000000..09c6b1ffd9c797
--- /dev/null
+++ b/clang/test/Modules/expose-static-inline-from-gmf-4.cppm
@@ -0,0 +1,40 @@
+// RUN: rm -rf %t
+// RUN: mkdir -p %t
+// RUN: split-file %s %t
+//
+// RUN: %clang -std=c++20 %t/a.cppm --precompile -o %t/a.pcm \
+// RUN: -DTEST_INLINE
+// RUN: %clang -std=c++20 %t/test.cc -fprebuilt-module-path=%t -fsyntax-only -Xclang -verify \
+// RUN: -DTEST_INLINE
+//
+// RUN: %clang -std=c++20 %t/a.cppm --precompile -o %t/a.pcm
+// RUN: %clang -std=c++20 %t/test.cc -fprebuilt-module-path=%t -fsyntax-only -Xclang -verify
+
+//--- a.h
+#ifdef TEST_INLINE
+#define INLINE inline
+#else
+#define INLINE
+#endif
+namespace ns {
+template <typename G> static void func() {}
+template <> INLINE void func<long>() {}
+template <typename T = long> void a() { func<T>(); }
+}
+
+//--- a.cppm
+module;
+#include "a.h"
+export module a;
+export using ns::a;
+
+//--- test.cc
+import a;
+auto m = (a(), 0);
+
+#ifdef TEST_INLINE
+// expected-no-diagnostics
+#else
+// [email protected]:9 {{no matching function for call to 'func'}}
+// [email protected]:2 {{in instantiation of function template specialization 'ns::a<long>' requested here}}
+#endif
diff --git a/clang/test/Modules/expose-static-inline-from-gmf-5.cppm b/clang/test/Modules/expose-static-inline-from-gmf-5.cppm
new file mode 100644
index 00000000000000..334af845a693dd
--- /dev/null
+++ b/clang/test/Modules/expose-static-inline-from-gmf-5.cppm
@@ -0,0 +1,26 @@
+// RUN: rm -rf %t
+// RUN: mkdir -p %t
+// RUN: split-file %s %t
+//
+// RUN: %clang -std=c++20 %t/a.cppm --precompile -o %t/a.pcm
+// RUN: %clang -std=c++20 %t/test.cc -fprebuilt-module-path=%t -fsyntax-only -Xclang -verify
+
+//--- a.h
+namespace ns {
+namespace {
+template <typename G> void func() {}
+}
+template <typename T = long> void a() { func<T>(); }
+}
+
+//--- a.cppm
+module;
+#include "a.h"
+export module a;
+export using ns::a;
+
+//--- test.cc
+import a;
+auto m = (a(), 0);
+
+// expected-no-diagnostics
|
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
ChuanqiXu9
approved these changes
Aug 20, 2024
clang/lib/Sema/SemaOverload.cpp
Outdated
Comment on lines
6937
to
6942
| /// Don't remove inline functions with internal linkage from the overload | ||
| /// set if they are declared in a GMF. | ||
| /// The global module is meant to be a transition mechanism for C and C++ | ||
| /// headers. | ||
| /// Inline functions with internal linkage are a common pattern in headers | ||
| /// to avoid ODR issues. |
There was a problem hiding this comment.
I'd like to make the comments to be more explicitly to mention the behavior is not strictly conforming the standard. Maybe we need to add a NOTE. BTW, maybe this worth a paper to WG21. WDYT?
There was a problem hiding this comment.
Sounds like a good idea! One direction could be to create an additional exception to basic.link#17 for exposure of inline TU-local entities declared in a GMF, so that the program would be no longer ill-formed in that case, right? depr.local's note would have to be changed to reflect this.
I will add to the comment that the proposed behavior violates basic.link#17.
There was a problem hiding this comment.
I feel the better wording will add a clause to https://eel.is/c++draft/basic.link#14. There are already some exceptions.
There was a problem hiding this comment.
Sorry for the delay! I have now updated the comment to include the fact that this fix/workaround violates [basic.link]p17, strictly speaking. I've also merged the main branch. On my machine, the (new) tests in ./tools/clang/test/Modules still pass. Does this work for you? Thanks again for the review!
Co-authored-by: Chuanqi Xu <[email protected]>
…functions-from-gmf
…functions-from-gmf
…functions-from-gmf
|
You may want to add a note in ReleaseNotes too. And if you want to write a paper to WG21, please CC me or I'd love to be a co-author with you. |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/65/builds/9952 Here is the relevant piece of the build log for the reference |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/94/builds/3306 Here is the relevant piece of the build log for the reference |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
In C, it is a common pattern to have
static inlinefunctions in headers to avoid ODR issues. Currently, when those headers are included in a GMF, the names are not found when two-phase name lookup and ADL is involved. Those names are removed bySema::AddOverloadCandidate.Similarly, in C++, sometimes people use templates with internal linkage in headers.
As the GMF was designed to be a transitional mechanism for headers, special case those functions in
Sema::AddOverloadCandidate.This fixes #98021.