On Thu, Apr 28, 2022 at 08:17:04PM +0100, Iain Sandoe wrote:
> Signed-off-by: Iain Sandoe <i...@sandoe.co.uk>
>
> PR c++/105426
>
> gcc/cp/ChangeLog:
>
> * coroutines.cc (register_local_var_uses): Allow promotion of unnamed
> temporaries to coroutine frame copies.
> ---
> gcc/cp/coroutines.cc | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/gcc/cp/coroutines.cc b/gcc/cp/coroutines.cc
> index 551ddc9cc41..2e393b2cddc 100644
> --- a/gcc/cp/coroutines.cc
> +++ b/gcc/cp/coroutines.cc
> @@ -3973,6 +3973,9 @@ register_local_var_uses (tree *stmt, int *do_subtree,
> void *d)
> else if (lvname != NULL_TREE)
> buf = xasprintf ("%s_%u_%u", IDENTIFIER_POINTER (lvname),
> lvd->nest_depth, lvd->bind_indx);
> + else
> + buf = xasprintf ("_D%u_%u_%u", DECL_UID (lvar), lvd->nest_depth,
> + lvd->bind_indx);
> /* TODO: Figure out if we should build a local type that has any
> excess alignment or size from the original decl. */
> if (buf)
This isn't going to play well with -fcompare-debug.
We don't guarantee same DECL_UID values between -g and -g0, just that
when two decls are created with both -g and -g0 they compare the same,
but with -g the gap in between them could be larger.
So names that include DECL_UID will be often different between -g and -g0.
Can't the FIELD_DECL be instead nameless? Say change coro_make_frame_entry
to do
tree id = name ? get_identifier (name) : NULL_TREE;
instead of
tree id = get_identifier (name);
?
Jakub
