[llvm-bugs] [Bug 43151] New: In Sema::SubstExceptionSpec: Assertion `ESI.Type != EST_Uninstantiated' failed

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=43151

            Bug ID: 43151
           Summary: In Sema::SubstExceptionSpec: Assertion `ESI.Type !=
                    EST_Uninstantiated' failed
           Product: clang
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: normal
          Priority: P
         Component: Frontend
          Assignee: unassignedclangbugs at nondot.org
          Reporter: aaronpuchert at alice-dsl.net
                CC: llvm-bugs at lists.llvm.org, neeilans at live.com,
                    richard-llvm at metafoo.co.uk

Created attachment 22445
  --> https://bugs.llvm.org/attachment.cgi?id=22445&action=edit
Small reproducer for the bug, compile with -std=c++17.

Compiling the following code with "-fsyntax-only -std=c++17" produces an
assertion failure:

template<typename T>                                                            
struct X {                                                                      
  using self = X<T>;                                                            

  template<bool noex>
  void f() noexcept(noex)
  {
    auto x = &self::f<noex>;
  }
};

void g()
{
  X<int> x;
  x.f<true>();
}

The trace shows:

1.      <eof> parser at end of file
2.      template-crash.cpp:6:8: instantiating function definition
'X<int>::f<true>'

The issue disappears when removing "-std=c++17", probably because the exception
specification is not part of the type then.

The function is called from clang::Sema::SubstituteExplicitTemplateArguments,
where we have EPI.ExceptionSpec.Type = clang::EST_Uninstantiated and
EPI.ExceptionSpec.SourceDecl = Function. On the other hand, the type of
EPI.ExceptionSpec.SourceTemplate has an ExceptionSpec with Type =
clang::EST_DependentNoexcept. So it seems we should be passing that one to
Sema::SubstExceptionSpec instead.

Indeed, the following seems to fix the assertion (and not introduce new
issues):

--- a/clang/lib/Sema/SemaTemplateDeduction.cpp
+++ b/clang/lib/Sema/SemaTemplateDeduction.cpp
@@ -3112,6 +3112,13 @@ Sema::SubstituteExplicitTemplateArguments(
     // so substitution into the type must also substitute into the exception
     // specification.
     SmallVector<QualType, 4> ExceptionStorage;
+    if (EPI.ExceptionSpec.Type == EST_Uninstantiated) {
+      assert(Function == EPI.ExceptionSpec.SourceDecl);
+      const auto *TplType = EPI.ExceptionSpec.SourceTemplate->getType()
+                                ->getAs<FunctionProtoType>();
+      EPI.ExceptionSpec = TplType->getExtProtoInfo().ExceptionSpec;
+      assert(EPI.ExceptionSpec.Type != EST_Uninstantiated);
+    }
     if (getLangOpts().CPlusPlus17 &&
         SubstExceptionSpec(
             Function->getLocation(), EPI.ExceptionSpec, ExceptionStorage,

But this seems like a hack, since in other instances the ExceptionSpec is
already instantiated correctly.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20190828/f77be45a/attachment.html>