[pkg-cryptsetup-devel] Bug#1122539: cryptsetup: autopkgtest 'upstream-testsuite' on amd64 is flaky

[Guilhem Moulin]

On Thu, 11 Dec 2025 at 22:29:41 +0100, Milan Broz wrote:
> On 12/11/25 12:30 PM, Luca Boccassi wrote:
>> The 'upstream-testsuite' is flaky on amd64 and often fails with:
>
> Is this CI running Debian unstable without any changes (udev etc)?

AFAICT it's an amd64 Debian testing kvm worker.  On Debian unstable (kernel 6.17.11+deb14-amd64)
the same test seems to consistently fail with

    $ sudo make -f Makefile.localtest -j tests CRYPTSETUP_PATH=/sbin TESTSUITE_NOSKIP=y
    […]
    [reencryption-compat-test]
    [1] Reencryption
    [2] Reencryption with data shift
    [3] Reencryption with keyfile
    [4] Encryption of not yet encrypted device
    [5] Reencryption using specific keyslot
    [6] Reencryption using all active keyslots
    [7] Reencryption of block devices with different block size
    [512 sector]Cannot use scsi_debug module (in use or compiled-in), test skipped.

Removing `TESTSUITE_NOSKIP=y` solves this one, but that kind of defeats the purpose of
the debci test as described in https://gitlab.com/cryptsetup/cryptsetup/-/merge_requests/256 .

> IOW can this be reproduced outside of your CI?

Looks like a race condition in `reencryption-compat-test`
https://gitlab.com/cryptsetup/cryptsetup/-/blob/v2.8.1/tests/reencryption-compat-test#L372 or
https://gitlab.com/cryptsetup/cryptsetup/-/blob/v2.8.1/tests/reencryption-compat-test#L114
(stages [5] and [512 sector]).

Tried it a few times locally and could only reproduce the latter, but
the former also works at times on the debci infrastructure (it's flaky).

I'll exclude that test for now.

-- 
Guilhem.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://alioth-lists.debian.net/pipermail/pkg-cryptsetup-devel/attachments/20251211/c2df4e5b/attachment.sig>