Image reports should not aggregate by test name

[Neil Williams]

When a single test produces multiple measurements, it is not possible to show the measurements in the image set.

https://staging.validation.linaro.org/dashboard/filters/~admin/TCP_STREAM-throughput

https://staging.validation.linaro.org/dashboard/image-reports/TCP_STREAM-throughput

Compared with:

http://validation.linaro.org/dashboard/image-reports/linaro-android-member-arm_vexpress-lt-kernel-ci

The netperf filter can be set either way - to specify multiple test cases under a single test or to list the same test repeatedly with a single test case each - neither case results in the data from the test cases actually being shown in the image set. The filter correctly shows the measurements (by listing by test_name:test_case) and the image set should do the same.

Note: this has nothing to do with MultiNode except that MultiNode was used to calculate the measurements.

[Neil Williams (codehelp)]

The core issue here is that the test image needs to execute a single command:
./netperf -t UDP_STREAM -H ${remote_ip} -- -m 1024 | ../../netperf/netperf2LAVA.py

However, this command produces multiple measurements in sequence:
https://staging.validation.linaro.org/scheduler/job/828/log_file#L_193_164

Each of these measurements needs to be visible as a line in the final graph.

The parsing already produces a unique test-case name but the test-name is outside the control of the parser.

[Stevan Radaković (stevanr)]

Thanks for the report.
However, we are in process of completely changing how benchmarks work on image reports (and how image reports work, for that matter). It's all described here https://cards.linaro.org/browse/LAVA-634 ... As you can see charting per test case will be possible here so this will not be an issue.
This was noticed on LCE13 already but we didn't take time to fix it because of the incoming changes...

[Stevan Radaković (stevanr)]

I suggest moving it to 'Wont fix'.