Report the connect status of the oldest connection in the socket pool.

net/socket/client_socket_pool_base_unittest.cc

Index: net/socket/client_socket_pool_base_unittest.cc
diff --git a/net/socket/client_socket_pool_base_unittest.cc b/net/socket/client_socket_pool_base_unittest.cc
index 391fb61e6d0557b28932f96059d13bf8db2b30ff..5cc94984a76d5a49183993dfa365c4b8af8fe1d1 100644
--- a/net/socket/client_socket_pool_base_unittest.cc
+++ b/net/socket/client_socket_pool_base_unittest.cc
@@ -1979,18 +1979,12 @@ TEST_F(ClientSocketPoolBaseTest, LoadStateTwoRequests) {
                     BoundNetLog());
   EXPECT_EQ(ERR_IO_PENDING, rv);
 
-  // If the first Job is in an earlier state than the second, the state of
-  // the second job should be used for both handles.
-  client_socket_factory_.SetJobLoadState(0, LOAD_STATE_RESOLVING_HOST);
-  EXPECT_EQ(LOAD_STATE_CONNECTING, handle.GetLoadState());
   EXPECT_EQ(LOAD_STATE_CONNECTING, handle2.GetLoadState());

[mmenke, 2015/04/20 14:49:16]

This seems a little weird now.  Suggest moving the
"client_socket_factory_.SetJobLoadState(0, LOAD_STATE_RESOLVING_HOST);" just
below the handle.Init / EXPECT_EQ lines, and getting rid of this check.

[mmenke, 2015/04/20 20:32:16]

Actually, going backwards in general seems a little weird, though the test was
going it before, too.  Maybe just move the first request forward instead (And in
a new test, move the second one forward).

[haavardm, 2015/04/21 13:28:36]

Done.

 
-  // If the second Job is in an earlier state than the second, the state of
-  // the first job should be used for both handles.
-  client_socket_factory_.SetJobLoadState(0, LOAD_STATE_SSL_HANDSHAKE);
-  // One request is farther
-  EXPECT_EQ(LOAD_STATE_SSL_HANDSHAKE, handle.GetLoadState());
-  EXPECT_EQ(LOAD_STATE_SSL_HANDSHAKE, handle2.GetLoadState());
+  // Check that both handles report the state of the first job.
+  client_socket_factory_.SetJobLoadState(0, LOAD_STATE_RESOLVING_HOST);
+  EXPECT_EQ(LOAD_STATE_RESOLVING_HOST, handle.GetLoadState());
+  EXPECT_EQ(LOAD_STATE_RESOLVING_HOST, handle2.GetLoadState());

[mmenke, 2015/04/20 14:49:16]

I'm paranoid:  Could you have another test just like this, but set the state of
the second socket (And have the second socket connect at the end)?

[haavardm, 2015/04/21 13:28:36]

Done.

 
   // Farthest along job connects and the first request gets the socket.  The
   // second handle switches to the state of the remaining ConnectJob.

[mmenke, 2015/04/20 14:49:16]

This comment is no longer accurate (It's the one that's not farthest
along...Maybe just "First job connects and...."

[haavardm, 2015/04/21 13:28:36]

Done.