Hi, Thank you both. Could you please send me a few examples? With shunit 2.1.8 I could reproduce the issue if I put significant load to my computer in the same time as I run the tests. (Twice failures from around 10 attempts.) Without load, or with shunit 2.1.7, it looks quite stable.
Regarding race conditions: So far I could not find any non-sequencial parts in the misspell-fixer's codebase. (Apart from using xargs with -P 4.) But I have not given up... :) Does sbuild do exactly the same as the live test agents? I am just wondering if that is a good enough test or I need something more if I find a fix. I suppose as a worst case, I could just disable the tests. But ideally I would like to fix them. Thank you. Best regards, Lajos On Thu, 29 Dec 2022, Lucas Nussbaum wrote: > Hi, > > Yes, every time. > > Maybe share your build log, and diff it with mine? -- Lajos Veres vla...@gmail.com 07927 460 778