Diego, I know you're working in that area, but the bug fix I'm working on is
kinda coming up to it too.
I realize you readded those two lines (in patch below) because of an ICE
(although I still think they don't make sense, i.e. adding the bindings of a
namespace to itself when they're already in there...). I checked and the only
reason this fixes your ICE is because it goes through and calls
varpool_finalize_decl for the names in that namespace bindings, the rest does
absolutely nothing...
I know you're restructuring this right now, but just in case you're keeping
some of that logic, I'm sharing my analysis on the matter now instead of
patching it later...
Also as highlighted in this patch, I don't think we need to validate that
NAMESPACE_LEVEL != NULL with the "if". It shouldn't be (I think... I even
tested it with an empty namespace...), if anything maybe we want to assert, but
not skip if it's NULL in my opinion.
Tested with bootstrap and pph regression testing.
Gab
2011-07-08 Gabriel Charette <gch...@google.com>
* gcc/cp/pph-streamer-in.c (pph_add_bindings_to_namespace):
NAMESPACE_LEVEL should never be null for a namespace, removed check.
diff --git a/gcc/cp/pph-streamer-in.c b/gcc/cp/pph-streamer-in.c
index d78ee91..55f7e12 100644
--- a/gcc/cp/pph-streamer-in.c
+++ b/gcc/cp/pph-streamer-in.c
@@ -1168,8 +1168,7 @@ pph_add_bindings_to_namespace (struct cp_binding_level
*bl, tree ns)
Preserve it. */
chain = DECL_CHAIN (t);
pushdecl_into_namespace (t, ns);
- if (NAMESPACE_LEVEL (t))
- pph_add_bindings_to_namespace (NAMESPACE_LEVEL (t), t);
+ pph_add_bindings_to_namespace (NAMESPACE_LEVEL (t), t);
}
}
--
This patch is available for review at http://codereview.appspot.com/4675069
