fix python3 py_test_histogram failure

[yonghong-song]

With python3 (specifically 3.6.3), py_test_histogram failed with:
Traceback (most recent call last):
File "./test_histogram.py", line 106, in test_multiple_key
bucket_sort_fn=bucket_sort)
File "/usr/lib/python3.6/site-packages/bcc/table.py", line 318, in print_log2_hist
buckets = bucket_sort_fn(buckets)
File "./test_histogram.py", line 98, in bucket_sort
buckets.sort()
AttributeError: 'dict_keys' object has no attribute 'sort'

The reason is for a dictionary tmp, tmp.keys() returns an object
of keys (a list) instead of the list itself for python3.
Using list(tmp.keys()) seems working for both python2 and python3.

Signed-off-by: Yonghong Song [email protected]

[goldshtn]

LGTM

…

On Mon, Apr 16, 2018 at 07:50 yonghong-song ***@***.***> wrote: @yonghong-song <https://github.com/yonghong-song> requested your review on: iovisor/bcc#1686 <#1686> fix python3 py_test_histogram failure as a code owner. — You are receiving this because your review was requested. Reply to this email directly, view it on GitHub <#1686 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABm3vSN7Ed1sfOoodBEFMLNM-Nh5dE-bks5tpCMWgaJpZM4TV3h3> .

[yonghong-song]

The test failure is fc27,

3: -------------------------------------------------------------------------------
3: test attach perf event
3: -------------------------------------------------------------------------------
3: /home/fedora/jenkins/workspace/bcc-pr/label/fc27/tests/cc/test_perf_event.cc:90
3: ...............................................................................
3: 
3: /home/fedora/jenkins/workspace/bcc-pr/label/fc27/tests/cc/test_perf_event.cc:134: FAILED:
3:   REQUIRE( counter.enabled >= 1000000000 )
3: with expansion:
3:   999605995 (0x3b94c6eb)
3:   >=
3:   1000000000 (0x3b9aca00)
3: 

The source code is

127 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 15, 0)
128   auto counter_table =
129       bpf.get_hash_table<int, struct bpf_perf_event_value>("counter");
130   auto counter = counter_table[0];
131   auto ret = bpf.get_hash_table<int, int>("ret");
132   REQUIRE(ret[0] == 0);
133   REQUIRE(counter.counter >= 0);
134   REQUIRE(counter.enabled >= 1000000000);
135   REQUIRE(counter.running >= 0);
136   REQUIRE(counter.running <= counter.enabled);
137 #endif

The test is only enabled for 4.15 and later, probably newer fc27.
The code to attach and detach perf event:

116   res = bpf.attach_perf_event(PERF_TYPE_SOFTWARE, PERF_COUNT_SW_CPU_CLOCK,
117                               "on_event", 0, 1000);
118   REQUIRE(res.code() == 0);
119   sleep(1);
120   res = bpf.detach_perf_event(PERF_TYPE_SOFTWARE, PERF_COUNT_SW_CPU_CLOCK);
121   REQUIRE(res.code() == 0);

For the failure below,

134   REQUIRE(counter.enabled >= 1000000000);

The VM counter read is slightly less

3:   999605995 (0x3b94c6eb)

It is possible that this is related to VM setup and how perf works for SW_CPU_CLOCK.
Will do an experiment separately.

Rerun the test for this patch.

[yonghong-song]

[buildbot, ok to test]