Control: tags -1 help Hi Santiago,
thanks for your QA work! well, I remember this test case being flaky since … always, at least since I've got the package… (I've never experienced a failure rate of 50% though, it was like once in a while, if I have to put a number on it, maybe 1 in 10 times) Still, the library is unchanged since 1.5 decades, (the only moving parts for this library are compiler versions…). For Sven's observation: Here, at my machine "magic numbers" for the random number to have a consistent failure doesn't work; also not when I encounter the failure, the next run is just fine, even using "my" random number… Bottom line, I still think this is not a library issue but a test suite issue - the library had seen in its best times a lots of usage and never reported an issue like that. However, I need to note that the "best times" had been already over when I took over that libary in 2012, upstream had already deprecated the library and had prominetnly on the homeage to use Kyoto Cabinet instead [1]. (And according to Wikipedia, this has been replaced with Tkrzz…) So I probably won't invest time in debugging this issue, as this feels like time should be put somewhere else… Instead I'll file bugs and try to convince the remaining reverse dependencies to migrate to something newer, possibly (if possible) with the goal to remove tokyocabinet from Debian.... (Therefore I've filed #1131879) [1] https://web.archive.org/web/20110310132255/https://fallabs.com/tokyocabinet/ -- tobi
