Test: I solemnly swear I tested this conflict resolution.
in cts-tf: run cts --module CtsNativeNetTestCases
Change-Id: I15253841993a79320dafa8669f0289c28f4ff4ea
This test verifies the socket address information is not leaked by
xt_qtaguid module inside the kernel. The sock_addr parsed from
proc/net/xt_qtaguid/ctrl file should always be 0
Bug: 62489688
Test: in cts-tf: run cts --package android.net.native
WiFi is not a CDD requirement so these tests should not fail when the
device under test does not have WiFi. The behavior is changed so that if
there is WiFi then both metered and unmetered tests will run. If there
is no WiFi and the current connection is metered then only metered tests
will run. If there is no WiFi and the current connection is not metered
then only unmetered tests will run.
Test: Successfully ran CTS test on both emulator and shamu.
BUG: 31648368
Change-Id: Ic643d2490e0a7e69b57a44599f1a4c57c67da873
Symptom: It should be more reasonable to control battery saver function from setting DB instead of plugging/unplugging charger for “CtsHostsideNetworkTests” test case.
Root Cause: The test function “setBatterySaverMode” of “CtsHostsideNetworkTests” use command to set setting DB when trying to turn on battery saver. But while trying to turn off battery saver, it only use charger plug-in event. It should be more reasonable to turn off battery saver through similar DB setting as this function did at turning on.
Solution: To control battery saver function from setting DB.
Project:
Note:
Test done by RD:
Futher testing need Q team's support:
Bug: 31897608
Change-Id: Id70ba458e85f98393d7652bb4e79bd182172c60f
Since Linux kernel 4.2, net.ipv6.auto_flowlabels is set by default, and
therefore the request and reply may have different IPv6 flow label.
Bug: 31444338
Test: On a kernel 4.4 board, run com.android.cts.net.HostsideNetworkTests#testVpn
Test: On a kernel 3.18 board, run echo 1 > /proc/sys/net/ipv6/auto_flowlabels, then
com.android.cts.net.HostsideNetworkTests#testVpn
Change-Id: I913bbf91574239a24cb32ae908834eb951ea2010
The behavior was not previous covered and broke in N
due to commit c80af6d8. There is an associated fix
in frameworks/base.
Some refactoring of existing tests to reduce
duplication and tidy up sockets after tests.
Test: Ran the new CTS test (before and after related fix)
Bug: 31205169
Change-Id: Ie94335bc535beefcc5301d5469de6b8211af0bab
- Retry on all cases (not only when expecting connected).
- Uses exponential back-off for timeout.
BUG: 27803922
Fixes: 30509643
Change-Id: I42454f43158598a72e30f290c27c5a02e80ea6d2
When counting all connectivity changed events, we were failing
on devices with sim cards because we were counting both wifi and
mobile connectivity changes the same. Since the test only
toggles wifi, changing the listener only to pay attention to wifi
changes.
Bug: 29346253
Change-Id: I1ed3b976bc21419218c780d4afc4a5e73f128496
Add parentheses around macro arguments used beside operators.
Use 'aoto' type to avoid clang-tidy warnings of missing parentheses
of 'TYPE *'.
Bug: 28705665
Change-Id: I5d2c3b2bdfb7775200f31a011758b9a35b14794d
If InetAddress.getAllByName("ipv6.google.com") throws
UnknownHostException, the test silently ignores it. This
causes a misleading failure, because the test then reuses the
addrs variable that is left over from the previous DNS query
for "www.google.com", and fails because it has an IPv4 address.
Fixes: 12210306
Bug: 29231261
Change-Id: I1dde945765d40a84eba139055306b07ebc97d0ec
