Hi! The following testcase ICEs, because due to the -frounding-math fold_const_call fails, which is it returns NULL, and returning NULL from cxx_eval* is wrong, all the callers rely on them to either return folded value or original with *non_constant_p = true.
The following patch does that, and additionally falls through into the default case where there is diagnostics for the !ctx->quiet case too. Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk? 2022-02-11 Jakub Jelinek <ja...@redhat.com> PR c++/104472 * constexpr.cc (cxx_eval_internal_function) <case IFN_VEC_CONVERT>: Only return fold_const_call result if it is non-NULL. Otherwise fall through into the default: case to return t, set *non_constant_p and emit diagnostics if needed. * g++.dg/cpp0x/constexpr-104472.C: New test. --- gcc/cp/constexpr.cc.jj 2022-02-08 20:15:33.997295271 +0100 +++ gcc/cp/constexpr.cc 2022-02-10 12:33:48.243321592 +0100 @@ -1840,13 +1840,10 @@ cxx_eval_internal_function (const conste false, non_constant_p, overflow_p); if (TREE_CODE (arg) == VECTOR_CST) - return fold_const_call (CFN_VEC_CONVERT, TREE_TYPE (t), arg); - else - { - *non_constant_p = true; - return t; - } + if (tree r = fold_const_call (CFN_VEC_CONVERT, TREE_TYPE (t), arg)) + return r; } + /* FALLTHRU */ default: if (!ctx->quiet) --- gcc/testsuite/g++.dg/cpp0x/constexpr-104472.C.jj 2022-02-10 12:44:09.869720657 +0100 +++ gcc/testsuite/g++.dg/cpp0x/constexpr-104472.C 2022-02-10 12:46:25.881838766 +0100 @@ -0,0 +1,9 @@ +// PR c++/104472 +// { dg-options "-O2 -frounding-math" } +// { dg-add-options float16 } +// { dg-require-effective-target float16 } + +typedef short __attribute__((__vector_size__ (16))) V; +typedef _Float16 __attribute__((__vector_size__ (16))) F; + +V v = __builtin_convertvector (__builtin_convertvector ((V){5534}, F), V) < 8; Jakub