TKO parser doesn't like network_WiFi_{AttenuatedPerf,Perf} throughput results |
||
Issue descriptionIn looking at new keyval entries in the context of this CL: https://chromium-review.googlesource.com/c/chromiumos/third_party/autotest/+/695707 We noticed that our existing keyval files for network_WiFi_AttenuatedPerf (and probably network_WiFi_Perf too) aren't getting parsed correctly. e.g.: https://storage.cloud.google.com/chromeos-autotest-results/146402154-chromeos-test/chromeos1-grover-host9/.parse.log?_ga=2.41266402.-1746072600.1506638659 complains about: ...tko parser: WARNING: line "ch001_modeHT40p_none_tcp_tx_atten066_throughput{perf}=143.91+-3.92" found in test iteration keyval could not be parsed ... This code is looking for 'float', not for 'float+-float' https://chromium.googlesource.com/chromiumos/third_party/autotest/+/master/tko/parsers/version_1.py#100 # Parse the actual value into a dict. try: if val_type == 'attr': attr_dict[key] = value elif val_type == 'perf': perf_dict[key] = float(value) else: raise ValueError except ValueError: msg = ('WARNING: line "%s" found in test ' 'iteration keyval could not be parsed') We should either improve the parser to handle +- (just drop the error bar, and report the mean netperf result?) or modify the keyval output in these tests, so it's parseable. Any thoughts? I'd kinda lean toward the former, since the error bars are useful in some other contexts (mostly manual inspection, so far).
,
Apr 2 2018
,
May 17 2018
I'm willing to fix up the parser, but what is this piece actually used for? We already have 2 different ways we write out results from network_WiFi_AttenuatedPerf: output_perf_value() and write_perf_keyval() The former logs to results-chart.json, seems to get parsed with the TKO parser, and somehow along the way ends up in chromeperf. But despite the above warnings, chromeperf seems to be working just fine. The latter logs to 'keyval' files and gets printed on stdout (which is nice) but then goes...nowhere? Am I missing something? I'd kinda like to know what/why I'm fixing, and I'd also like to avoid reporting the same results too many different ways.
,
May 25 2018
Issue 846510 has been merged into this issue. |
||
►
Sign in to add a comment |
||
Comment 1 by briannorris@chromium.org
, Oct 5 2017