I was planning to apply this as obvious - but it is needed for the
next patch to be posted - so noting here now. I discussed with one
of the original coroutines paper authors the idea that, if the original
function was marked noexcept, then we should carry that onto the
outlined body. This was not the design intent, so it is quite expected
that the ramp might be noexcept - but that a resumer via a handle could
still receive an exception. Given this and the fact we unconditionally
add exceptions code to the body, I consider this patch to be obvious and
the EH code certainly gets very confused without it (fun to debug).
Any comments?
thanks
Iain
--- 8<---
We must flag that the resumer might throw (since the wrapping of the
original function body unconditionally adds a try-catch/rethrow). We
also add code that might throw - even when the original function body
would not.
gcc/cp/ChangeLog:
* coroutines.cc (build_actor_fn): Set can_throw.
Signed-off-by: Iain Sandoe <i...@sandoe.co.uk>
---
gcc/cp/coroutines.cc | 3 +++
1 file changed, 3 insertions(+)
diff --git a/gcc/cp/coroutines.cc b/gcc/cp/coroutines.cc
index 18c0a4812c4..d82fa46f208 100644
--- a/gcc/cp/coroutines.cc
+++ b/gcc/cp/coroutines.cc
@@ -2412,6 +2412,9 @@ build_actor_fn (location_t loc, tree coro_frame_type,
tree actor, tree fnbody,
bool spf = start_preparsed_function (actor, NULL_TREE, SF_PRE_PARSED);
gcc_checking_assert (spf);
gcc_checking_assert (cfun && current_function_decl && TREE_STATIC (actor));
+ if (flag_exceptions)
+ /* We, unconditionally, add a try/catch and rethrow. */
+ cp_function_chain->can_throw = true;
tree stmt = begin_function_body ();
tree actor_bind = build3 (BIND_EXPR, void_type_node, NULL, NULL, NULL);
--
2.39.2 (Apple Git-143)
