On Sat, 2016-12-24 at 08:34 +0800, Ed Greshko wrote:
On 12/19/16 06:22, Patrick O'Callaghan wrote:
> Of course, but in this case there is clearly a database lock for some
> unexplained reason, which the dnf authors(s) did not expect to ever
> happen (or they would have caught the exception).
>
> I guess I should BZ it.
Just a side note which *may* be relevant.
I had never see the "sqlite3.OperationalError: database is locked" error until
today. The
only thing that was different was that in testing things related to the "Most
system
update requires system reboot or session restart" thread I enabled the
"dnf-plugin-tracer"
plugin. Do you happen to have that enabled on your system ?
Yes, I always enable it. It's a plausible theory that this could be responsible.
If so, it may be best to BZ the akmod package. Also, the error may
only happen if there
is a large enough number of packages being updated causing the tracer process to take a
long enough time.
It seems unlikely that a specific package could be the cause of the bug
rather than dnf itself (including the plugin of course). Package
installation is often going to run scripts (complex ones in the case of
akmod) so it would be up to dnf to handle the locking correctly.
poc