On 05.02.2024 00:50, PLD th-i686 builder wrote: > test.spec (cryptsetup-debug): FAILED ... > + test cafa743b687f1581b71a20ebfc068f9fe12653e945c417311b861c9217e66adf '=' > 06e2f159d6d541169da75222a7f0bedd0fc700502d63809f577b48f41eace335
Something concerning is going on i686/x32 builders. All started with cryptsetup tests failing there without any clear root cause. Code involved in failing test is pretty straightforward and has no issues as far as I can tell. Managed to narrow it down to: - happens only on PLD builders - happens only on 32-bit archs - happens only when direct io (O_DIRECT) is involved The concerning part is that writing with O_DIRECT produces non-sense result. Parts of write might be missing or they might be scattered in different places even though they should be contiguous. Simple reproducer is currently present in test package on cryptsetup-debug branch. I've tried to replicate the env (i686 VM, XFS) but without success. Not sure if it's block device driver? XFS? or something else? Can't do much more without ability to reproduce. Any ideas are welcome. _______________________________________________ pld-devel-en mailing list pld-devel-en@lists.pld-linux.org http://lists.pld-linux.org/mailman/listinfo/pld-devel-en