[Rpm-maint] [rpm-software-management/rpm] Check the integrity of the RPM database and system file consistency when reboot during an update process (Issue #3947)
Fellipe Henrique
notifications at github.com
Tue Sep 16 14:10:13 UTC 2025
fhbash created an issue (rpm-software-management/rpm#3947)
This issue comes from this [DNF5 issue](https://github.com/rpm-software-management/dnf5/issues/2437) thread, we believe this will be better to be handled on RPM instead of DNF5, copying Petr comment:
>
> I think you can replace the "reboot" with "kill -9 THE_DNF5_PROCESS".
>
> While I understand the motivation, checking for DNF5-did-not-finish-a-trasaction would be only a partial fix. You can get the RPM transaction unfinished in many other ways, e.g. when executing rpm command without DNF5. You would also need implement the check in libdn5. Not in DNF5 program, because packit or dnf5-daemon uses that.
>
> For all the reasons I believe the right place for the implementation should be in RPM (librpm), not in libdnf5: RPM when asked for starting a transaction would recognized that it did not finished a previous transaction and returned a dedicated error. DNF5 (libdnf5) would than simply forward that error to the user.
>
Thanks!
**Background**
If the system restarts suddenly while the dnf command is updating the system, then upon the next system boot, dnf can continue executing a new system update operation. However, due to the previously aborted transaction, this is very likely to lead to entries for different versions of the same package name existing within the RPM database. Also files on the system may be corrupted.
Please provide the package NVR for which the bug is seen:
**How reproducible is this bug?:**
Should be easy.
**Steps to reproduce**
Install Fedora 42+.
While dnf is performing transaction for updating the system, reboot the system.
**Expected results**
After reboot, the next run of dnf should stop and remind user to verify the system's RPM data and files' state.
Actual results
After reboot, the dnf could continue to run another transaction with corrupted RPM database.
--
Reply to this email directly or view it on GitHub:
https://github.com/rpm-software-management/rpm/issues/3947
You are receiving this because you are subscribed to this thread.
Message ID: <rpm-software-management/rpm/issues/3947 at github.com>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rpm.org/pipermail/rpm-maint/attachments/20250916/372de787/attachment.htm>
More information about the Rpm-maint
mailing list