[clang] Use CPlusPlus language option instead of Bool #80975
Conversation
As it was pointed out in llvm#80724, we should not be checking `getLangOpts().Bool` when determining something related to logical operators, since it only indicates that bool keyword is present, not which semantic logical operators have. As a side effect a missing `-Wpointer-bool-conversion` in OpenCL C was restored since like C23, OpenCL C has bool keyword but logical operators still return int.
llvmbot
added
clang
|
@llvm/pr-subscribers-clang Author: Mariya Podchishchaeva (Fznamznon) ChangesAs it was pointed out in #80724, we should not be checking Full diff: https://github.com/llvm/llvm-project/pull/80975.diff 2 Files Affected:
diff --git a/clang/lib/Sema/SemaChecking.cpp b/clang/lib/Sema/SemaChecking.cpp
index b071a02ca3713f..6588aa23e1116f 100644
--- a/clang/lib/Sema/SemaChecking.cpp
+++ b/clang/lib/Sema/SemaChecking.cpp
@@ -16129,10 +16129,10 @@ static void CheckConditionalOperator(Sema &S, AbstractConditionalOperator *E,
/// Check conversion of given expression to boolean.
/// Input argument E is a logical expression.
static void CheckBoolLikeConversion(Sema &S, Expr *E, SourceLocation CC) {
- // While C23 does have bool as a keyword, we still need to run the bool-like
- // conversion checks as bools are still not used as the return type from
- // "boolean" operators or as the input type for conditional operators.
- if (S.getLangOpts().Bool && !S.getLangOpts().C23)
+ // Run the bool-like conversion checks only for C since there bools are
+ // still not used as the return type from "boolean" operators or as the input
+ // type for conditional operators.
+ if (S.getLangOpts().CPlusPlus)
return;
if (E->IgnoreParenImpCasts()->getType()->isAtomicType())
return;
diff --git a/clang/test/SemaOpenCL/operators.cl b/clang/test/SemaOpenCL/operators.cl
index cf359acd5acb97..76a7692a7105c8 100644
--- a/clang/test/SemaOpenCL/operators.cl
+++ b/clang/test/SemaOpenCL/operators.cl
@@ -118,6 +118,6 @@ kernel void pointer_ops(){
bool b = !p;
b = p==0;
int i;
- b = !&i;
+ b = !&i; // expected-warning {{address of 'i' will always evaluate to 'true'}}
b = &i==(int *)1;
}
|
|
Not sure about what to put in a release note, and whether it is necessary at all. |
There was a problem hiding this comment.
The change in OpenCL behavior looks reasonable to me, but let's wait for @AnastasiaStulova to have a chance to weigh in.
I think we can skip a release note as this is almost an NFC change except for the OpenCL bits. We could put a release note under OpenCL for this if we wanted though.
The changes themselves LGTM barring feedback on OpenCL.
|
@svenvh , I wonder if you could provide some feedback for OpenCL bits here? |
Thanks! |
As it was pointed out in #80724, we should not be checking
getLangOpts().Boolwhen determining something related to logical operators, since it only indicates that bool keyword is present, not which semantic logical operators have. As a side effect a missing-Wpointer-bool-conversionin OpenCL C was restored since like C23, OpenCL C has bool keyword but logical operators still return int.