W dniu 28.09.2021 o 09:32, Julian Sikorski pisze:
W dniu 22.09.2021 o 21:07, Julian Sikorski pisze:
> Am 22.09.21 um 19:38 schrieb Fabio Valentini:
>> On Wed, Sep 22, 2021 at 6:35 PM Julian Sikorski <belegdol(a)gmail.com>
>> wrote:
>>>
>>> W dniu 21.09.2021 o 23:12, Richard W.M. Jones pisze:
>>>> On Tue, Sep 21, 2021 at 10:16:17PM +0200, Julian Sikorski wrote:
>>>>> W dniu 21.09.2021 o 11:00, Richard W.M. Jones pisze:
>>>>>> On Mon, Sep 20, 2021 at 11:45:39AM -0600, Jerry James wrote:
>>>>>>> On Mon, Sep 20, 2021 at 10:49 AM Julian Sikorski
>>>>>>> <belegdol(a)gmail.com> wrote:
>>>>>>>> my local kernel rebuilds have started failing for no
apparent
>>>>>>>> reason - I
>>>>>>>> was using a similar command successfully for several
months.
>>>>>>>> /mnt/openmediavault is a samba share. This is what gets
output
>>>>>>>> into the log:
>>>>>>>
>>>>>>> [snip]
>>>>>>>
>>>>>>>>
/mnt/openmediavault/kernel/results/fedora-34-x86_64/pesign-113-16.fc35/pesign-113-16.fc34.x86_64.rpm:
>>>>>>>>
>>>>>>>> (39, fsync failed: Permission denied)
>>>>>>>
>>>>>>> I suspect that ^^^^ is the real problem, and the incorrect
>>>>>>> checksum is
>>>>>>> a result of not being able to read or write the filesystem.
Can
>>>>>>> you
>>>>>>> verify correct ownership and permissions on every directory
in that
>>>>>>> path?
>>>>>>
>>>>>> If fsync(2) failed then that would be happening after the file
>>>>>> descriptor was open, so it wouldn't be filesystem permissions
but
>>>>>> probably SELinux.
>>>>>>
>>>>>> Rich.
>>>>>>
>>>>>
>>>>> I am seeing no errors in setroubleshoot and setting selinux to
>>>>> permissive does not help either. Can it be the selinux of the
>>>>> bootstrapped instance, and if so, how can I control it? There is
>>>>> nothing obvious in the logs of the host:
>>>>
>>>> AFAIK mock only ever uses one kernel (the host) so disabling SELinux
>>>> on the host rules out my SELinux theory.
>>>>
>>>> Rich.
>>>>
>>> This is all super strange. If I delete the pesign result dir, it is
>>> built without issues. Moreover, repodata folder is also created, with
>>> the sha256 checksum in the file matching the one of the pesign rpm.
>>> What
>>> is odd though, is that even after mock fails, gedit is claiming that
>>> repomd.xml and the data file extracted from primary.xml.gz keep
>>> changing
>>> on disk. Can it be some odd system clock issue between my desktop an
>>> my NAS?
>>
>> Oh my, that filesystem is mounted over a network? What filesystem is
>> backing that directory? SMB? NFS?
>> I assume that the issue you're seeing is caused by something like "NFS
>> doesn't support the fsync call properly" ...
>>
>> Fabio
>>
> It is a local network samba share, from a host running armbian and
> openmediavault. This used to work until last week or so, which is why
> I am starting to suspect some strange clock problem. I have tried
> rebooting the NAS but it did not help.
>
> Best regards,
> Julian
I have now managed to generate logs on the NAS, see attached. There are
two errors which stand out: NT_STATUS_NO_EAS_ON_FILE and
NT_STATUS_ACCESS_DENIED later. Does this give any indication as to what
might be happening?
Best regards,
Julian
I have been investigating this further. When building from a different
client, I am getting most interesting results:
$ mock --chain --localrepo=/mnt/openmediavault/kernel -r
fedora-34-x86_64 goffice/goffice-0.10.49-1.fc36.src.rpm
gnumeric/gnumeric-1.12.49-1.fc36.src.rpm
will keep working from the laptop as goffice-0.10.49 was never rebuilt
from the desktop, as long as goffice-0.10.50 results folder is not
present in /mnt/openmediavault/kernel.
$ mock --chain --localrepo=/mnt/openmediavault/kernel -r
fedora-34-x86_64 goffice/goffice-0.10.50-2.fc36.src.rpm
gnumeric/gnumeric-1.12.50-2.fc36.src.rpm
will fail when run from the laptop, even if the goffice-0.10.50-2.fc36
folder does not exist when the build is started. Moreover, building to a
different folder does not work either:
$ mock --chain --localrepo=/mnt/openmediavault/kernel1 -r
fedora-34-x86_64 goffice/goffice-0.10.50-2.fc36.src.rpm
gnumeric/gnumeric-1.12.50-2.fc36.src.rpm
This is beyond strange. To me it looks as if either the permissions set
from the desktop PC are somehow wrong and additionally persist the
folder deletion, or as if createrepo needs different permissions to
generate checksum for goffice-0.10.50-2 than for goffice-0.10.49-1.
Neither of the above makes too much sense.
What does createrepo use to generate the checksum exactly? It would be
good to have a simpler reproducer.
Best regards,
Julian
Best regards,
Julian