(I was attempted to push this via our obvious role, and IIRC Richi
pre-approved such changes a few years ago. At the same time I'm
wondering why nobody else picked this up earlier?)
We currently issue a good deal of warnings of the following kind:
In file included from /scratch/tmp/gerald/GCC-HEAD/gcc/coretypes.h:500:
.../GCC-HEAD/gcc/poly-int.h:378:1: warning: 'poly_int' defined as a
struct template here but previously declared as a class template; this
is valid, but may result in linker errors under the Microsoft C++ ABI
[-Wmismatched-tags]
378 | struct poly_int
| ^
.../GCC-HEAD/gcc/poly-int.h:32:38: note: did you mean struct here?
32 | template<unsigned int N, typename T> class poly_int;
| ^~~~~
| struct
A `grep "'poly_int' defined as a struct template" | cut -d: -f1 | uniq -c`
shows 749 issue for a typical bootstrap.
Addressing this brings down the bootstrap log by 8.5% - from 80454 lines
down to 73613.
Gerald
---
gcc/poly-int.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/gcc/poly-int.h b/gcc/poly-int.h
index 77f78752e74..d99f63b671f 100644
--- a/gcc/poly-int.h
+++ b/gcc/poly-int.h
@@ -374,7 +374,7 @@ template<> struct poly_int_fullness<true> { using type =
poly_int_full; };
/* A class containing polynomial integers. The polynomial has N coefficients
of type C, and N - 1 indeterminates. */
template<unsigned int N, typename C>
-struct poly_int
+class poly_int
{
public:
poly_int () = default;
--
2.51.1
