Hi! The following testcase ICEs because we haven't verified if reduction decl has reference type that TREE_TYPE of the reference is a complete type, require_complete_type on the decl doesn't ensure that.
Bootstrapped/regtested on x86_64-linux and i686-linux, committed to trunk. 2021-07-21 Jakub Jelinek <ja...@redhat.com> PR c++/101516 * semantics.c (finish_omp_reduction_clause): Also call complete_type_or_else and return true if it fails. * g++.dg/gomp/pr101516.C: New test. --- gcc/cp/semantics.c.jj 2021-07-15 10:16:12.972581906 +0200 +++ gcc/cp/semantics.c 2021-07-20 13:31:04.972039268 +0200 @@ -6070,7 +6070,8 @@ finish_omp_reduction_clause (tree c, boo if (!processing_template_decl) { t = require_complete_type (t); - if (t == error_mark_node) + if (t == error_mark_node + || !complete_type_or_else (oatype, NULL_TREE)) return true; tree size = size_binop (EXACT_DIV_EXPR, TYPE_SIZE_UNIT (oatype), TYPE_SIZE_UNIT (type)); --- gcc/testsuite/g++.dg/gomp/pr101516.C.jj 2021-07-20 13:51:41.690789542 +0200 +++ gcc/testsuite/g++.dg/gomp/pr101516.C 2021-07-20 13:51:16.463141545 +0200 @@ -0,0 +1,8 @@ +// PR c++/101516 + +void +foo (int (&v) []) +{ + #pragma omp parallel reduction (+:v) // { dg-error "invalid use of array with unspecified bounds" } + ; +} Jakub