Hello,
the case I encountered the original issue with was a partially broken RPM
database, which allowed the "Computing transaction" to work, but failed at
"Committing transaction" when installing a package.
To reproduce it, the RPM database can be broken intentionally:
# echo foo > /var/lib/rpm/Arch
Then try to install some package...
# smart install|reinstall|upgrade <package>
SmartPM output & exit code before the fix (commit
1dc5f5d5c844585eec114be9480e0e4d8e60d09c):
Loading cache...
...
Computing transaction...
...
Committing transaction...
Preparing... ######################################## [ 0%]
1:Installing <package> ######################################## [100%]
Output from <package>-<version>@<arch>:
rpmdb: BDB0004 fop_read_meta: /var/lib/rpm/Arch: unexpected file type or format
error: cannot open Arch(1022) index: Invalid argument(22)
DB: Berkeley DB 6.0.30: (January 23, 2014)
Saving cache...
# echo $?
0
And after the fix:
Loading cache...
...
Computing transaction...
...
Committing transaction...
Preparing... ######################################## [ 0%]
1:Installing <package> ######################################## [100%]
Output from <package>-<version>@<arch>:
rpmdb: BDB0004 fop_read_meta: /var/lib/rpm/Arch: unexpected file type or format
error: cannot open Arch(1022) index: Invalid argument(22)
DB: Berkeley DB 6.0.30: (January 23, 2014)
error: Unknown error
# echo $?
1
I've only ever tested this with breaking the Arch index database; perhaps the
behaviour will be different when breaking another index, and most likely RPM
will fail earlier when the main /var/lib/rpm/Packages database is broken.
Either way, perhaps this can be used for a test case.
Regards,
Daniel
--
_______________________________________________
Openembedded-core mailing list
[email protected]
http://lists.openembedded.org/mailman/listinfo/openembedded-core
