On 3/16/26 4:02 PM, Marek Polacek wrote:
On Mon, Mar 16, 2026 at 03:02:36PM -0400, Jason Merrill wrote:
On 3/16/26 10:40 AM, Marek Polacek wrote:
Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk?

-- >8 --
When printing the diagnostics in this test, we're saying
"constexprstd::meta::info dm" which misses a space.  We
should use pp_cxx_ws_string which does pp_c_maybe_whitespace
when printing "std::meta::info".

        PR c++/124489

gcc/cp/ChangeLog:

        * error.cc (dump_type) <case NULLPTR_TYPE>: Use pp_cxx_ws_string

Switch cases in ChangeLogs are generally spelled with [], e.g.

* error.cc (dump_type) [NULLPTR_TYPE]

OK with that change.

Huh.  I've always used the <case foo> style and looking at our ChangeLogs
I'm not the only one.  Same in binutils-gdb.  Even rms's emacs ChangeLog
uses that style:
1998-01-02 Richard Stallman <[email protected]>

        * print.c (strout): Take args SIZE and SIZE_BYTE;
        operate on both chars and bytes.
        (print_string): Pass new arg to strout.
        If not using strout, fetch a whole multibyte char at once.
        (write_string): Pass new arg to strout.
        (write_string_1): Likewise.
        (print) <case Lisp_String>: Scan by chars and bytes.
        (print) <case Lisp_Symbol>: Scan name by chars and bytes.

I've assumed that [] is for #ifdefs, as documented at
<https://www.gnu.org/prep/standards/html_node/Conditional-Changes.html>.
But <https://www.gnu.org/prep/standards/html_node/Style-of-Change-Logs.html>
says nothing about switch cases.  Neither does
<https://gcc.gnu.org/codingconventions.html#ChangeLogs>.

git_commit.py suggests that <case foo> is expected:

                 # Strip everything that is not a filename in "line":
                 # entities "(NAME)", cases "<PATTERN>", conditions
                 # "[COND]", entry text (the colon, if present, and
                 # anything that follows it).

I can use either but we should state somewhere what the right format is.

Hunh, interesting. I see the [] pattern from other people going back to ChangeLog-2000; I guess I got it from them rather than the docs.

So, OK without change then.

        instead of pp_string.
        <case META_TYPE>: Likewise.

gcc/testsuite/ChangeLog:

        * g++.dg/reflect/diag6.C: New test.
---
   gcc/cp/error.cc                      |  4 ++--
   gcc/testsuite/g++.dg/reflect/diag6.C | 10 ++++++++++
   2 files changed, 12 insertions(+), 2 deletions(-)
   create mode 100644 gcc/testsuite/g++.dg/reflect/diag6.C

diff --git a/gcc/cp/error.cc b/gcc/cp/error.cc
index 25e438c1784..f198a1fe065 100644
--- a/gcc/cp/error.cc
+++ b/gcc/cp/error.cc
@@ -875,11 +875,11 @@ dump_type (cxx_pretty_printer *pp, tree t, int flags)
         break;
       case NULLPTR_TYPE:
-      pp_string (pp, "std::nullptr_t");
+      pp_cxx_ws_string (pp, "std::nullptr_t");
         break;
       case META_TYPE:
-      pp_string (pp, "std::meta::info");
+      pp_cxx_ws_string (pp, "std::meta::info");
         break;
       case SPLICE_SCOPE:
diff --git a/gcc/testsuite/g++.dg/reflect/diag6.C 
b/gcc/testsuite/g++.dg/reflect/diag6.C
new file mode 100644
index 00000000000..61caa082551
--- /dev/null
+++ b/gcc/testsuite/g++.dg/reflect/diag6.C
@@ -0,0 +1,10 @@
+// PR c++/124489
+// { dg-do compile { target c++26 } }
+// { dg-additional-options "-freflection" }
+
+void
+g ()
+{
+  constexpr decltype(^^::) dm = ^^int; // { dg-message ".constexpr std::meta::info 
dm. previously declared here" }
+  constexpr decltype(^^::) dm = ^^int; // { dg-error "redeclaration of .constexpr 
std::meta::info dm." }
+}

base-commit: 64f95a0eeaf1cb6d07cc414c9b2953a494e03be1


Marek


Reply via email to