From 493c15113e5b832a21a81a36a18bc6e6cff2e2f3 Mon Sep 17 00:00:00 2001
From: Craig Tiller <ctiller@google.com>
Date: Wed, 27 Jan 2016 11:55:44 -0800
Subject: [PATCH] Add comment
---
src/core/client_config/subchannel.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/src/core/client_config/subchannel.c b/src/core/client_config/subchannel.c
index e39b7fcf32..748eef9bed 100644
--- a/src/core/client_config/subchannel.c
+++ b/src/core/client_config/subchannel.c
@@ -519,6 +519,10 @@ static void publish_transport(grpc_exec_ctx *exec_ctx, grpc_subchannel *c) {
}
/* publish */
+ /* TODO(ctiller): this full barrier seems to clear up a TSAN failure.
+ I'd have expected the rel_cas below to be enough, but
+ seemingly it's not.
+ Re-evaluate if we really need this. */
gpr_atm_full_barrier();
GPR_ASSERT(gpr_atm_rel_cas(&c->connected_subchannel, 0, (gpr_atm)con));
c->connecting = 0;
--
GitLab