2026-07-29 06:26:17 <ikke> cryptsetup has a test failure on loongarch, but it's unclear what's failing. Anyone has time to take a look at it?
2026-07-29 06:26:19 <ikke> https://build.alpinelinux.org/buildlogs/build-edge-loongarch64/main/cryptsetup/cryptsetup-2.8.7-r0.log
2026-07-29 06:41:12 <huajingyun> Yeah, I've noticed it and will take a look at it later
2026-07-29 06:51:59 <mps> it passed on CI
2026-07-29 06:54:20 <mps> is the same kernel on CI and builder
2026-07-29 08:42:06 <huajingyun> mps: aha, thanks for info
2026-07-29 08:42:37 <huajingyun> I just ran `abuild rootbld` on one machine with kernel 6.18.29-0, and it passed without any issues
2026-07-29 08:43:16 <huajingyun> but now the same test error is reproducible on another machine
2026-07-29 09:42:35 <huajingyun> The difference appears to be with twofish-ecb. In CI log, twofish-ecb is shown as [N/A]. according to ./tcrypt-compat-test  behavior, when twofish-ecb is [N/A], it deletes certain test files including tcrypt-images/tc_3-ripemd160-xts-aes-twofish, so the tests are skipped
2026-07-29 09:45:13 <huajingyun> In the builder's error logs, twofish-ecb is shown as [OK], so it proceeds to test tcrypt-images/tc_3-ripemd160-xts-aes-twofish, and then fails
2026-07-29 09:47:23 <ikke> What determines whether twofish-ecb is available?
2026-07-29 09:47:46 <huajingyun> I'm guessing that running `grep twofish /proc/crypto` on the CI and the Builder would yield different results
2026-07-29 09:48:04 <huajingyun> could you take a look?
2026-07-29 09:48:19 <ikke> Yes, in a bit
2026-07-29 09:49:41 <huajingyun> Thanks
2026-07-29 09:50:03 <huajingyun> Perhaps the builder's output contains something like ecb(twofish) while the CI's does not
2026-07-29 09:59:19 <huajingyun> If there is indeed a difference between the two, that would be a bit strange, since mps mentioned that both (CI and builder) are running the same kernel
2026-07-29 09:59:35 <huajingyun> anyway, ikke, I'm going to leave for now and will check back later
2026-07-29 10:25:04 <ikke> Builder:
2026-07-29 10:25:06 <ikke> name         : cbc(twofish)
2026-07-29 10:25:18 <ikke> name         : ecb(twofish)
2026-07-29 10:26:16 <ikke> On both CI hosts, it's not available
2026-07-29 10:27:36 <ikke> The twofish module was loaded on the builder, but not on the CI hosts
2026-07-29 10:37:58 <mps> try disable (blacklist) it on builder
2026-07-29 10:38:31 <mps> changelog says it could be problem
2026-07-29 12:32:27 <ikke> Test passed after blacklisting
2026-07-29 13:21:46 <mps> good \o/
2026-07-29 13:40:45 <ikke> So I suppose in lxc, it's able to load new kernel modules, while in docker it's not