> Am 23.05.2025 um 16:49 schrieb Thomas Schwinge <tschwi...@baylibre.com>:
>
> This fell out of me looking into PR119835. This doesn't resolve the
> underlying
> issue, but instead of failing GIMPLE semantics verification just by chance in
> the 'GIMPLE pass: nrv' context, it makes the issue observable generally.
> (... thereby regressing a small number of offloading test cases where host vs.
> offload compilers disagree on 'aggregate_value_p' for functions that return
> aggregate types.)
>
> This cross-references just the three places in the code that I ran into;
> likely there are more?
>
> No regressions for powerpc64le-unknown-linux-gnu, x86_64-pc-linux-gnu
> bootstrap
> and 'make check' (without offloading configured).
I think this is a step in the wrong direction in absence of quoting the wrong
thing that happens downstream when we violate this (an assert does not
qualify). ESP. When at the same time we allow the actual thing returned to be
a register (aka SSA name)
Richard
> PR middle-end/119835
> gcc/
> * tree-cfg.cc (verify_gimple_return): Verify 'GIMPLE_RETURN' vs.
> 'RESULT_DECL' if 'aggregate_value_p'.
> * gimplify.cc (gimplify_return_expr): Cross-reference it.
> * tree-nrv.cc (pass_nrv::execute): Likewise.
> ---
> gcc/gimplify.cc | 1 +
> gcc/tree-cfg.cc | 23 +++++++++++++++++++++++
> gcc/tree-nrv.cc | 1 +
> 3 files changed, 25 insertions(+)
>
> diff --git a/gcc/gimplify.cc b/gcc/gimplify.cc
> index 4f385b1b779..ac296ab70bf 100644
> --- a/gcc/gimplify.cc
> +++ b/gcc/gimplify.cc
> @@ -1903,6 +1903,7 @@ gimplify_return_expr (tree stmt, gimple_seq *pre_p)
> }
>
> /* If aggregate_value_p is true, then we can return the bare RESULT_DECL.
> + This is checked in 'tree-cfg.cc:verify_gimple_return'.
> Recall that aggregate_value_p is FALSE for any aggregate type that is
> returned in registers. If we're returning values in registers, then
> we don't want to extend the lifetime of the RESULT_DECL, particularly
> diff --git a/gcc/tree-cfg.cc b/gcc/tree-cfg.cc
> index 6a95b82ff40..2fc99d54934 100644
> --- a/gcc/tree-cfg.cc
> +++ b/gcc/tree-cfg.cc
> @@ -4892,6 +4892,29 @@ verify_gimple_return (greturn *stmt)
> return true;
> }
>
> + /* Verify 'gimplify.cc:gimplify_return_expr' property:
> + "If aggregate_value_p is true, then we can return the bare RESULT_DECL."
> + */
> + tree decl_result = DECL_RESULT (current_function_decl);
> + if (aggregate_value_p (decl_result, current_function_decl))
> + {
> + tree op_check;
> + if (TREE_CODE (op) == SSA_NAME)
> + op_check = SSA_NAME_VAR (op);
> + else
> + op_check = op;
> + if (op_check != decl_result)
> + {
> + error ("operand in return statement differs from %qs",
> + get_tree_code_name (TREE_CODE (decl_result)));
> + if (op != op_check)
> + debug_generic_stmt (op);
> + debug_generic_stmt (op_check);
> + debug_generic_stmt (decl_result);
> + return true;
> + }
> + }
> +
> if ((TREE_CODE (op) == RESULT_DECL
> && DECL_BY_REFERENCE (op))
> || (TREE_CODE (op) == SSA_NAME
> diff --git a/gcc/tree-nrv.cc b/gcc/tree-nrv.cc
> index 180ce39de4c..6ac77567600 100644
> --- a/gcc/tree-nrv.cc
> +++ b/gcc/tree-nrv.cc
> @@ -174,6 +174,7 @@ pass_nrv::execute (function *fun)
> /* In a function with an aggregate return value, the
> gimplifier has changed all non-empty RETURN_EXPRs to
> return the RESULT_DECL. */
> + /* See also 'tree-cfg.cc:verify_gimple_return'. */
> ret_val = gimple_return_retval (return_stmt);
> if (ret_val)
> gcc_assert (ret_val == result);
> --
> 2.34.1
>
