On 1/22/25 12:26 AM, Philipp Kern wrote: > On 1/21/25 5:49 PM, Chris Hofstaedtler wrote: >> pyacoustid currently fails its autopkgtest on s390x, and thus does >> not migrate back to testing. It also prevents eartag from re-enterting >> testing. >> >> Issues preventing migration: >> ∙ ∙ autopkgtest for pyacoustid/1.3.0-1: amd64: Pass, arm64: Pass, armel: >> Pass, armhf: Pass, i386: Pass, riscv64: Pass, s390x: Regression or new test ♻ >> >> https://ci.debian.net/packages/p/pyacoustid/testing/s390x/55645697/ >> https://ci.debian.net/packages/p/pyacoustid/testing/s390x/ > > I suspect python3-audioread. Looking at "for block in pcmiter" in > acoustid.fingerprint() I get many blocks with 8k on s390x but blocks > with variable length on amd64. The metadata (length, channels, sampling > rate) seems correct, though. > > The beginning of the first block on s390x: > > 1d001d001c001c001e001e002100210022002200210021002000200020002000 > > On amd64: > > 1e001e001c001c001f001f002100210022002300210021002100210020002000 > > That's super weird though.
And also the code supports falling back to fpcalc for this - which works correctly (meaning that chromaprint is likely fine). But if audioread (and chromaprint) is present it will use that. Kind regards Philipp Kern
