Created attachment 36034 [details] Source code and system info The program itself has a bug, in function f the expression "size(A(1,:))" at x variable definition, line 11, should to be "size(B(1,:))" instead. But that causes a segmentation fault on compiling.
Confirmed from 4.8 up to trunk (6.0), backtrace * thread #1: tid = 0x3763a43, 0x000000010098b77a f951`get_chain_decl at tree-nested.c:240, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x50) frame #0: 0x000000010098b77a f951`get_chain_decl at tree-nested.c:240 237 static tree 238 get_frame_type (struct nesting_info *info) 239 { -> 240 tree type = info->frame_type; 241 if (!type) 242 { 243 char *name; (lldb) bt * thread #1: tid = 0x3763a43, 0x000000010098b77a f951`get_chain_decl at tree-nested.c:240, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x50) * frame #0: 0x000000010098b77a f951`get_chain_decl at tree-nested.c:240 frame #1: 0x000000010098b77a f951`get_chain_decl(info=0x0000000144000b30) + 10 frame #2: 0x000000010098dc53 f951`get_nonlocal_debug_decl(info=0x0000000144000b30, decl=0x000000014350b990) + 723 at tree-nested.c:858 frame #3: 0x000000010098dd79 f951`convert_nonlocal_reference_op(tp=0x00000001436d6108, walk_subtrees=<unavailable>, data=0x00007fff5fbfeef0) + 265 at tree-nested.c:941 frame #4: 0x0000000100b3e6a3 f951`walk_tree_1(tp=0x00000001436d6108, func=0x000000010098dc70, data=0x00007fff5fbfeef0, pset=0x0000000000000000, lh=0x0000000000000000)(tree_node**, int*, void*), void*, hash_set<tree_node*, default_hash_traits<tree_node*> >*, tree_node* (*)(tree_node**, int*, tree_node* (*)(tree_node**, int*, void*), void*, hash_set<tree_node*, default_hash_traits<tree_node*> >*)) + 371 at tree.c:11296 frame #5: 0x000000010098de38 f951`convert_nonlocal_reference_op(tp=0x00000001436d6108, walk_subtrees=<unavailable>, data=0x00007fff5fbfeef0) + 456 at tree-nested.c:1037 frame #6: 0x0000000100b3e6a3 f951`walk_tree_1(tp=0x00000001436d4d18, func=0x000000010098dc70, data=0x00007fff5fbfeef0, pset=0x0000000000000000, lh=0x0000000000000000)(tree_node**, int*, void*), void*, hash_set<tree_node*, default_hash_traits<tree_node*> >*, tree_node* (*)(tree_node**, int*, tree_node* (*)(tree_node**, int*, void*), void*, hash_set<tree_node*, default_hash_traits<tree_node*> >*)) + 371 at tree.c:11296 frame #7: 0x0000000100686063 f951`walk_gimple_op(stmt=0x00000001436d4cd0, callback_op=0x000000010098dc70, wi=0x00007fff5fbfeef0)(tree_node**, int*, void*), walk_stmt_info*) + 1539 at gimple-walk.c:204 frame #8: 0x00000001006863d4 f951`walk_gimple_stmt(gsi=0x00007fff5fbfeab0, callback_stmt=0x000000010098e560, callback_op=0x000000010098dc70, wi=0x00007fff5fbfeef0)(gimple_stmt_iterator*, bool*, walk_stmt_info*), tree_node* (*)(tree_node**, int*, void*), walk_stmt_info*) + 148 at gimple-walk.c:562 frame #9: 0x0000000100686621 f951`walk_gimple_seq_mod(pseq=<unavailable>, callback_stmt=0x000000010098e560, callback_op=0x000000010098dc70, wi=0x00007fff5fbfeef0)(gimple_stmt_iterator*, bool*, walk_stmt_info*), tree_node* (*)(tree_node**, int*, void*), walk_stmt_info*) + 81 at gimple-walk.c:54 frame #10: 0x0000000100686543 f951`walk_gimple_stmt(gsi=0x00007fff5fbfeb40, callback_stmt=0x000000010098e560, callback_op=0x000000010098dc70, wi=0x00007fff5fbfeef0)(gimple_stmt_iterator*, bool*, walk_stmt_info*), tree_node* (*)(tree_node**, int*, void*), walk_stmt_info*) + 515 at gimple-walk.c:608 frame #11: 0x0000000100686621 f951`walk_gimple_seq_mod(pseq=<unavailable>, callback_stmt=0x000000010098e560, callback_op=0x000000010098dc70, wi=0x00007fff5fbfeef0)(gimple_stmt_iterator*, bool*, walk_stmt_info*), tree_node* (*)(tree_node**, int*, void*), walk_stmt_info*) + 81 at gimple-walk.c:54 frame #12: 0x0000000100686513 f951`walk_gimple_stmt(gsi=0x00007fff5fbfebd0, callback_stmt=0x000000010098e560, callback_op=0x000000010098dc70, wi=0x00007fff5fbfeef0)(gimple_stmt_iterator*, bool*, walk_stmt_info*), tree_node* (*)(tree_node**, int*, void*), walk_stmt_info*) + 467 at gimple-walk.c:572
I believe the code is invalid. Here a small rewrite program bug implicit none real, target, allocatable :: a(:,:) allocate(a(10,10)) a = 1.0 print*, f(a) contains function f(b) result(x) real, target, intent(in):: b(:,:) real, pointer :: c(:,:) real :: x(size(c(1,:))) c => b(1:2,:) x = c(2,:) end function f end program bug The line 'real :: x(size(c(1,:))' is referencing an unassociated pointer. If I add the 'integer :: n = size(c(1,:)) I get % gfc6 -c bug.f90 bug.f90:15:29: integer :: n = size(c(1,:)) 1 Error: Deferred array 'c' at (1) is not permitted in an initialization expression A similar error should probably occur here.
*** Bug 86120 has been marked as a duplicate of this bug. ***
We might need to tighten the check of restricted expressions: diff --git a/gcc/fortran/expr.cc b/gcc/fortran/expr.cc index 663fe63dea6..6996e1e93d1 100644 --- a/gcc/fortran/expr.cc +++ b/gcc/fortran/expr.cc @@ -2849,6 +2852,21 @@ check_inquiry (gfc_expr *e, int not_restricted) && ap->expr->expr_type == EXPR_VARIABLE && asym->attr.dummy && asym->attr.optional) return MATCH_NO; + + if (not_restricted == 0 + && ap->expr->expr_type == EXPR_VARIABLE + && !(asym->attr.flavor == FL_PARAMETER) + && !asym->attr.dummy + && !asym->attr.in_common + && !asym->attr.use_assoc + && !asym->attr.host_assoc + && !asym->attr.implied_index + && (asym->attr.allocatable + || asym->attr.pointer + || (asym->as && asym->as->type == AS_DEFERRED))) + { + return MATCH_ERROR; + } } return MATCH_YES; This needs to be refined further.