On 7/30/26 3:00 PM, Patrick Palka wrote:
Changes in v2:
- Outright revert r16-5967, and instead fix PR119343 more directly
for sake of backports.
Bootstrapped and regtested on x86_64-pc-linux-gnu, does this look
OK for trunk/16?
-- >8 --
In r16-5967-gbae0ed69e1862a we removed the mark_used call from
resolve_nondeduced_context under the rationale that it should be
the callers' responsiblity to do so.c
Removing this call however now means that resolve_nondeduced_context
could return a specialization whose type is not yet fully resolved
(i.e. has an uninstantiated noexcept or undeduced return type), and
callers that immediately inspect TREE_TYPE of the result (such as
standard_conversion and build_conditional_expr) now misbehave.
In light of such callers, this patch reverts r16-5967 and more
directly fixes the PR119343 bug by just propagating error_mark_node
from resolve_nondeduced_context during convert_to_void.
@@ -1573,7 +1573,8 @@ convert_to_void (tree expr, impl_conv_void implicit,
tsubst_flags_t complain)
default:;
}
expr = resolve_nondeduced_context (expr, complain);
- if (!mark_single_function (expr, complain))
+ if (expr == error_mark_node
+ || !mark_single_function (expr, complain))
I still don't understand why this makes a difference, given that
mark_single_function has
if (expr == error_mark_node)
return false;
so for expr == error_mark_node it should return false, and we should
already return error_mark_node?
Jason
