We had a testfailure in sequoia on s390x in Debian, and this reproduces the error.

Test output:

363s failures:
363s
363s ---- packet::test::mutate_eq_discriminates stdout ----
363s thread 'packet::test::mutate_eq_discriminates' panicked at '[quickcheck] TEST FAILED. Arguments: (-----BEGIN PGP ARMORED FILE-----
363s
363s w8BRBU4CAuyZ1F8AIADMWALvqVxDf787kQH3ZhXdebbRpneHZjl7+KlY2wfTVEad
363s 7wAJB43JggBr6F9Yprbu48JX+xFAFkXbhh/AWD1W7qQXlKIz4HDd3BBE2QNfzKVN
363s GQtfXs4AdI53W1L/KibaFcT0cwD8ANn/wtxjP6FO+2eGvRoc6wG+ogDwa8FlrAFy
363s hoH9r0Bu3zoXex0jBmRtgN8HXj4A/05z76OfuCyZhgHC/9/8AU+6hyTUYOzyVPES
363s uYttawHWpFxW/i9zAUmk0P8ZyRE+/8CCGr8wyscORUuCVrCjjWzSlEpDnh1aNJIY
363s /99JYLLapQcL2G7BkCrSKQsB5NkJ+Nb/6JpiAAE+aVYARRal
363s =49fD
363s -----END PGP ARMORED FILE-----
363s , 1863593116114256088)', /usr/share/cargo/registry/quickcheck-1.0.3/src/tester.rs:165:28
363s stack backtrace:
363s    0: std::panicking::begin_panic
363s              at /usr/src/rustc-1.70.0/library/std/src/panicking.rs:610:12
363s    1: quickcheck::tester::QuickCheck::quickcheck
363s              at /usr/share/cargo/registry/quickcheck-1.0.3/src/tester.rs:165:28
363s    2: quickcheck::tester::quickcheck
363s              at /usr/share/cargo/registry/quickcheck-1.0.3/src/tester.rs:183:5
363s    3: sequoia_openpgp::packet::test::mutate_eq_discriminates
363s              at /usr/share/cargo/registry/quickcheck-1.0.3/src/lib.rs:62:21
363s    4: sequoia_openpgp::packet::test::mutate_eq_discriminates::{{closure}}
363s              at /usr/share/cargo/registry/quickcheck-1.0.3/src/lib.rs:58:31
363s    5: core::ops::function::FnOnce::call_once
363s              at /usr/src/rustc-1.70.0/library/core/src/ops/function.rs:250:5
363s    6: core::ops::function::FnOnce::call_once
363s              at /usr/src/rustc-1.70.0/library/core/src/ops/function.rs:250:5
363s note: Some details are omitted, run with  for a verbose backtrace.
363s
363s
363s failures:
363s     packet::test::mutate_eq_discriminates
363s

Note: This MR introduces a test failure, so it's more for documentation purposes so that we don't lose track of it.

It wasn't obvious to me on how to solve it, if it represents a real weakness in the PartialEq implementation or not.