Revert "Merge "resolve merge conflicts of a97c792 to kitkat-cts-dev" into kitkat-cts-dev"
This reverts commit 9feaaaf446, reversing
changes made to 65fbe1fcd153c5647f8d1e1a7443b3f04b7b25fd.
(cherry-picked from master: 0d4434d998
conflict + added Log)
For testTrafficStatsForLocalhost's UID testing, it will also calcuate the wlan0 interface.
There are some TCP re-tranmission in SSLCertificateSocketFactoryTest and it is the same UID as this test case. Need to consider those extra pacetks.
For example,
Before testTrafficStatsForLocalhost test casae:
01-08 15:49:11.316 7826 7839 D TrafficStats: parseUidStats, buffer: 14 wlan0 0x0 10067 0 31857 67 4582 55 31857 67 0 0 0 0 4582 55 0 0 0 0
01-08 15:49:11.335 7826 7839 D TrafficStats: parseUidStats, buffer: 24 lo 0x0 10067 0 40 1 60 1 40 1 0 0 0 0 60 1 0 0 0 0
After testTrafficStatsForLocalhost test casae:
01-08 15:49:19.210 7826 7839 D TrafficStats: parseUidStats, buffer: 14 wlan0 0x0 10067 0 31857 67 4738 58 31857 67 0 0 0 0 4738 58 0 0 0 0
01-08 15:49:19.212 7826 7839 D TrafficStats: parseUidStats, buffer: 24 lo 0x0 10067 0 1155336 2053 1155408 2054 1155336 2053 0 0 0 0 1155408 2054 0 0 0 0
=> There are three extra IP packets after testing due to TCP FIN retreamsisions in previous test case
Suggest to add some extra IP packets for consider the extra traffic in wlan0 interface.
Signed-off-by: Junjie Hu <junjie.hu@mediatek.com>
Change-Id: I981f98fc8647469fb105361516b6a59c53530c70
Noticed as a result of 8fbc09bde8.
One of the AF_UNIX sockets being created in three tests was not
being closed (until GC/finalization). Because the name is now
used in multiple tests some of the tests would fail in the CTS
test runner ("Address already in use") if no GC had occurred.
Bug: 3106438
Change-Id: I7debe5ed26d9d09e71a3d8a1b5b7d85b13e6e069
This fixes the following cases for us:
android.net.cts.ConnectivityManagerTest#testGetAllNetworkInfo
android.net.cts.ConnectivityManagerTest#testGetNetworkInfo
android.net.cts.ConnectivityManagerTest#testIsNetworkSupported
android.net.cts.ConnectivityManagerTest#testRequestRouteToHost
android.net.cts.ConnectivityManagerTest#testSetNetworkPreference
We need to use the prebuilt binary but that has dependency to
private resource identities that differs in our environment. With
this change the resources are looked up in run-time to avoid the
build time dependency.
Change-Id: I6579338b591ca7a0da3f03f796136269c7789780
(cherry picked from commit b29370db34)
The testcase was manipulating the Wifi / Cell networks,
presumably in an effort to make data pass over each type of
network.
The test was sending / receiving data using the loopback
interface so it doesn't make much sense.
testExecute_withMobile would fail if the device did not
have a SIM / mobile data connection *and* if the wifi
network was up when the test ran; it would fail waiting
for the mobile data network to enter the CONNECTED state.
If the wifi connection was down at the start it would
assume the mobile data connection was up already and
proceed with the test (which passed, because it
didn't actually use the network).
This change chops out parts of this test to leave just
HttpClient-related code.
Bug: 21874093
Change-Id: I8248b840f58f939661eff2345f70a9e786e593f3
It is a issue caused by bad test case.
Please ensure the previous vpn connection is disconnected before running next
vpn test case, this will affect the test result.And the CTS test case also
tries to connect socket before the vpn rule is set in netd.
It is the same issue as http://b/18436087 .
The delay 300ms is too short, and recommend extend to 1000 or 3000 ms.
Since L MR1 is known as supporting some low memory device, we suggest fixing
the new CTS test case for some low-end devices.
Change-Id: I6051c95c663e867fc66c580ce8bd8d25fce0b8fc
Signed-off-by: Junjie Hu <junjie.hu@mediatek.com>
In current implementation it's been assumed that there will be only one
interface registered for service discovery in mdnsresponder. In reality
there's possibility that some real but currently unused interfaces will
have link-local address assigned prior to registration, which will cause
mdnsresponder to register these interfaces for service discovery. This
will result in SERVICE_LOST and SERVICE_FOUND events to be received more
than one time. Current test implementation is not 100% prone for this
situation - depending upon time gap between both events it can PASS or
FAIL.
Fix by removing an assert on number of received events.
Change-Id: I147dc4d600f41f6d63d2b9e4bcb10efe90d5b3ec
(cherry picked from commit 970cbe09b5)
If one sets a *SendBufferSize* value which is less than
(2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff))) then the value
retrieved will be (2 * (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))).
Otherwise, the value will be doubled.
Modify the test to consider both cases.
Change-Id: I2d89393b1ad441783c8a082b072a806b8901830c
Signed-off-by: Constantin Musca <constantin.musca@intel.com>
The test is failing because it expects ::1" to resolve to
"localhost", but it now resolves to "ip6-localhost".
Bug: 18791191
Bug: 19953381
Change-Id: I62279b6219c32e817253c686bf408f0a2fc1fa6b
Revert "DO NOT MERGE ANYWHERE."
This reverts commit b7c798f9b84c7c100aa37e6aa77e7f31460c7bd7, reversing
changes made to 4e8ecc32db5d3a17ccec099681425c34b630d23b.
