[clang] Change "bad" to "unsupported" in register type error

[DavidSpickett]

This is maybe a personal take but I expect "bad" to either mean:

• Allowed but not ideal, like a "bad" memory alignment might work but it is slow.
• The tool won't allow it but is going to tell me why it didn't.

The current error doesn't elaborate so I think it's best we just say "unsupported" instead. This is clear that the type used is not allowed at all.

[llvmbot]

@llvm/pr-subscribers-clang

Author: David Spickett (DavidSpickett)

Changes

This is maybe a personal take but I expect "bad" to either mean:

• Allowed but not ideal, like a "bad" memory alignment might work but it is slow.
• The tool won't allow it but is going to tell me why it didn't.

The current error doesn't elaborate so I think it's best we just say "unsupported" instead. This is clear that the type used is not allowed at all.

---

Full diff: https://github.com/llvm/llvm-project/pull/111550.diff

4 Files Affected:

• (modified) clang/include/clang/Basic/DiagnosticSemaKinds.td (+2-1)
• (modified) clang/lib/Sema/SemaDecl.cpp (+2-1)
• (modified) clang/test/Sema/asm.c (+2-2)
• (modified) clang/test/Sema/caret-diags-register-variable.cpp (+1-1)

diff --git a/clang/include/clang/Basic/DiagnosticSemaKinds.td b/clang/include/clang/Basic/DiagnosticSemaKinds.td
index 583475327c5227..82588cea4155c1 100644
--- a/clang/include/clang/Basic/DiagnosticSemaKinds.td
+++ b/clang/include/clang/Basic/DiagnosticSemaKinds.td
@@ -9386,7 +9386,8 @@ let CategoryName = "Inline Assembly Issue" in {
     "global register variables on this target">;
   def err_asm_register_size_mismatch : Error<"size of register '%0' does not "
     "match variable size">;
-  def err_asm_bad_register_type : Error<"bad type for named register variable">;
+  def err_asm_unsupported_register_type : Error<
+    "unsupported type for named register variable">;
   def err_asm_invalid_input_size : Error<
     "invalid input size for constraint '%0'">;
   def err_asm_invalid_output_size : Error<
diff --git a/clang/lib/Sema/SemaDecl.cpp b/clang/lib/Sema/SemaDecl.cpp
index 83d71913f8635e..072f43d360ee1c 100644
--- a/clang/lib/Sema/SemaDecl.cpp
+++ b/clang/lib/Sema/SemaDecl.cpp
@@ -7961,7 +7961,8 @@ NamedDecl *Sema::ActOnVariableDeclarator(
       }
 
       if (!R->isIntegralType(Context) && !R->isPointerType()) {
-        Diag(TInfo->getTypeLoc().getBeginLoc(), diag::err_asm_bad_register_type)
+        Diag(TInfo->getTypeLoc().getBeginLoc(),
+             diag::err_asm_unsupported_register_type)
             << TInfo->getTypeLoc().getSourceRange();
         NewVD->setInvalidDecl(true);
       }
diff --git a/clang/test/Sema/asm.c b/clang/test/Sema/asm.c
index 630a5e85dd9131..6cd95c71604d44 100644
--- a/clang/test/Sema/asm.c
+++ b/clang/test/Sema/asm.c
@@ -191,8 +191,8 @@ void iOutputConstraint(int x){
 struct foo {
   int a;
 };
-register struct foo bar asm("esp"); // expected-error {{bad type for named register variable}}
-register float baz asm("esp"); // expected-error {{bad type for named register variable}}
+register struct foo bar asm("esp"); // expected-error {{unsupported type for named register variable}}
+register float baz asm("esp"); // expected-error {{unsupported type for named register variable}}
 
 register int r0 asm ("edi"); // expected-error {{register 'edi' unsuitable for global register variables on this target}}
 register long long r1 asm ("esp"); // expected-error {{size of register 'esp' does not match variable size}}
diff --git a/clang/test/Sema/caret-diags-register-variable.cpp b/clang/test/Sema/caret-diags-register-variable.cpp
index 24f5061d4b4d2c..c2d2fbe0c581ae 100644
--- a/clang/test/Sema/caret-diags-register-variable.cpp
+++ b/clang/test/Sema/caret-diags-register-variable.cpp
@@ -4,7 +4,7 @@ struct foo {
   int a;
 };
 
-//CHECK: {{.*}}: error: bad type for named register variable
+//CHECK: {{.*}}: error: unsupported type for named register variable
 //CHECK-NEXT: {{^}}register struct foo bar asm("esp");
 //CHECK-NEXT: {{^}}         ^~~~~~~~~~{{$}}
 register struct foo bar asm("esp");

[Sirraide]

LGTM

(I’ve definitely seen ‘bad X’ in other places before, and I’ve personally never really had a problem w/ that sort of wording, but I also can’t deny that ‘unsupported X’ is unquestionably clearer ;Þ)

[Sirraide]

I think we also might want to add a release note for this since this is technically improving a diagnostic, if only in a very minor way... I’m not sure if we have a clear-cut policy for that when it comes to changes that only change the wording of a diagnostic.

[DavidSpickett]

Added a release note.