Re: [PATCH RFC] Fix ICE due to shared BLOCK node in coroutine generation [PR103328]

[Jason Merrill <jason@redhat.com>]

On 3/17/22 07:37, Benno Evers via Gcc-patches wrote:
> The coroutine transformation moves the original function body into a
> newly created actor function, but the block of the
> `current_binding_level` still points into the original function,
> causing the block to be shared between the two functions if it is
> subsequently used. This may cause havoc later on, as subsequent
> compiler passes for one function will also implicitly modify the
> other. See https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103328#c19 for
> a more detailed writeup.
> 
> This patch fixes the issue locally, but I'm not familiar with the GCC
> code base so I'm not sure if this is the right way to go about it or
> if there's a cleaner way to reset the current binding level. If this
> is the way to go I'm happy to extend the patch with a testcase and
> changelog entry.

Please do, it looks like a good fix to me.  Iain, does it make sense to 
you as well?

> ---
>   gcc/cp/coroutines.cc | 2 ++
>   1 file changed, 2 insertions(+)
> 
> diff --git a/gcc/cp/coroutines.cc b/gcc/cp/coroutines.cc
> index b1bfdc767a4..eb5f80f499b 100644
> --- a/gcc/cp/coroutines.cc
> +++ b/gcc/cp/coroutines.cc
> @@ -4541,6 +4541,8 @@ morph_fn_to_coro (tree orig, tree *resumer, tree
> *destroyer)
>     BLOCK_VARS (top_block) = BIND_EXPR_VARS (ramp_bind);
>     BLOCK_SUBBLOCKS (top_block) = NULL_TREE;
> 
> +  current_binding_level->blocks = top_block;
> +
>     /* The decl_expr for the coro frame pointer, initialize to zero so that we
>        can pass it to the IFN_CO_FRAME (since there's no way to pass a type,
>        directly apparently).  This avoids a "used uninitialized" warning.  */