[clang] nonblocking/nonallocating attributes (was: nolock/noalloc) (PR #84983)
Doug Wyatt via cfe-commits
cfe-commits at lists.llvm.org
Mon May 27 10:17:37 PDT 2024
================
@@ -7525,25 +7525,21 @@ static Attr *getCCTypeAttr(ASTContext &Ctx, ParsedAttr &Attr) {
std::optional<FunctionEffectMode>
Sema::ActOnEffectExpression(Expr *CondExpr, StringRef AttributeName) {
- auto BadExpr = [&]() {
- Diag(CondExpr->getExprLoc(), diag::err_attribute_argument_type)
- << ("'" + AttributeName.str() + "'") << AANT_ArgumentIntegerConstant
- << CondExpr->getSourceRange();
+ if (DiagnoseUnexpandedParameterPack(CondExpr))
return std::nullopt;
- };
-
- if (CondExpr->isTypeDependent() || CondExpr->isValueDependent()) {
- if (CondExpr->containsUnexpandedParameterPack())
- return BadExpr();
+ if (CondExpr->isTypeDependent() || CondExpr->isValueDependent())
return FunctionEffectMode::Dependent;
- }
std::optional<llvm::APSInt> ConditionValue =
CondExpr->getIntegerConstantExpr(Context);
- if (!ConditionValue)
- return BadExpr();
- return ConditionValue->getExtValue() ? FunctionEffectMode::True
- : FunctionEffectMode::False;
+ if (!ConditionValue) {
+ Diag(CondExpr->getExprLoc(), diag::err_attribute_argument_type)
+ << ("'" + AttributeName.str() + "'") << AANT_ArgumentIntegerConstant
----------------
dougsonos wrote:
The difficulty here is:
- this is a pre-existing diagnostic; I didn't see a need for a new one just for quotes
- the other users of the pre-existing diagnostic use `<<` with a `ParsedAttr`, which apparently quotes the attribute (?), but this function can be called from `TreeTransform` (for template instantiation), where there is only a `FunctionEffect` (whose name is passed in).
And then there are usages like:
```
Diag(AttrLoc, diag::err_attribute_argument_type)
<< "vector_size" << AANT_ArgumentIntegerConstant
<< SizeExpr->getSourceRange();
```
, passing a string constant; it would appear that in this case the attribute isn't quoted, which seems inconsistent.
https://github.com/llvm/llvm-project/pull/84983
More information about the cfe-commits
mailing list