[12:00:50] (03CR) 10Awight: [C: 03+1] "Looks right! A next step could be to wrap the parallel steps up as a ParallelNpmInstall command, to clean up cmd.py ." [integration/quibble] - 10https://gerrit.wikimedia.org/r/754866 (owner: 10Kosta Harlan) [12:02:23] hashar: +1 I like the suggestion to expand fullrun slightly. [12:04:20] kostajh: Thanks for the debugging! So what's missing is the two stdout messages. Seems important ;-) [12:05:05] Even important enough to block the whole patch, I don't want the stdout capture to randomly fail when we need it. [12:08:13] hmm, I was having trouble with this exact code under pytest and even had to skip for that reason. pytest was taking over the std* streams and they were replaced with an internal thing which couldn't `dup2`. But that resulted in a crash rather than missing loglines. [12:37:45] k, let me know if there's something I can do to help debug. [12:38:02] it doesn't seem like a random failure, though, it looks like a macOS / pytest issue. [12:38:28] The stdout capture seems to work if I run the code via the application in normal use (not in pytest) [12:50:09] hrm [12:50:53] Okay well if it's just pytest, maybe I'll step back my rash "blocky block" talk [20:19:28] 10Quibble, 10Continuous-Integration-Infrastructure, 10Wikidata, 10Patch-For-Review, 10Release-Engineering-Team (CI & Testing services): Terminating MySQL takes several minutes in (Wikibase?) CI jobs - https://phabricator.wikimedia.org/T265615 (10hashar)