updating vitest fails randomness tests

[felixroos]

updating from 0.21.1 to 0.24.5 changes the results of the randomness tests. it seems that one of both versions does not reset its state or something. eventually related with vitest-dev/vitest#1754 but not really sure.

it seems that the newer version does not behave like the repl would, e.g. http://strudel.tidalcycles.org/#c3RhY2soCiAgICBzKCJiZCA8c2QgY3A%2BIikKICAgIC5kZWxheSgiPDAgLjU%2BIikKICAgIC5kZWxheXRpbWUoIi4xNiB8IC4zMyIpCiAgICAuZGVsYXlmZWVkYmFjaygiLjYgfCAuOCIpCiAgKS5zb21ldGltZXMoeD0%2BeC5zcGVlZCgiLTEiKSkK always starts with .8 delayfeedback so does the current test. with the new vitest version, it starts with .6 ..

tried various combinations of --no-isolate and --no-threads flags, but no luck so far.

[yaxu]

Hmm so is randomness not a deterministic pure function of time in strudel?

[felixroos]

yes it is, but between different versions of the test runner, the determinism seems to have switched how it always behaves :)

[yaxu]

heh strange !

[felixroos]

the problem should be that the mini notation "|" alignment as well as "?" operator use a stateful randomness, which behaves differently when multiple patterns are queried one after the other. see 306004e i don't remember what the seed approach originally tried to solve, maybe @bpow still knows?

[felixroos]

i don't remember what the seed approach originally tried to solve, maybe @bpow still knows?

probably this: #165 (comment)