On Saturday, January 31, 2015 10:16:01 PM David Goodenough wrote: > Unfortunately I have a problem with the KDE bugs system. It says my login > is wrong, won't send me the password reminder, and won't let me create a > new one. No idea why it does not like me. > > David > > On Saturday 31 January 2015 22:54:32 you wrote: > > tag 775990 -moreinfo > > thanks > > > > Hey, > > > > thanks for the further information. But please if you have some > > information > > to bugs, attach them to the bugreport. It is much easier to follow a > > specific discussion instead of switching between different lists... > > > > did you checked if there is a matching upstream bug? This one should go > > upsteam. Can you check for a matching bug (bugs.kde.org) - if there is > > none > > create a new one. Afterwards please let us know the bugnr. > > > > regads, > > > > sandro > > > > ---------- Weitergeleitete Nachricht ---------- > > > > Betreff: Re: Sqlite akonadi transaction mode fix > > Datum: Montag, 26. Januar 2015, 09:45:03 > > Von: David Goodenough <david.goodeno...@btconnect.com> > > An: debian-...@lists.debian.org > > > > On Wednesday 21 January 2015 09:49:18 David Goodenough wrote: > > > I use KdePIM and thus akonadi with SqLite. I am running unstable and am > > > up to date with fixes. > > > > > > Looking in the KDE bug database I found:- > > > > > > http://osdir.com/ml/kde-commits/2014-07/msg02228.html > > > > > > which seems to address the repeated dialog boxes I get while filters run > > > on > > > my mailboxes, and the log messages of which:- > > > > > > DATABASE ERROR: > > > Error code: 6 > > > DB error: "database is deadlocked" > > > Error text: "database is deadlocked Unable to fetch row" > > > Query: "UPDATE PimItemTable SET atime = ? WHERE ( ( PimItemTable.id = > > > ? > > > ) > > > > > > )" > > > > > > is an example. All the errors I get currently are deadlock errors. > > > > > > I don't think this fix is in the current build, and I realise that > > > Debian > > > is getting ready for a new release and thus it is unlikely that this fix > > > will be able to be slotted in any time soon. Can I simply get the > > > source, add this one patch and rebuild, or is it more complicated. Also > > > which package(s) would I have to rebuild? Is it just > > > akonadi-backend-sqlite? > > > > > > David > > > > Well having loaded the source and looked at it, this patch is already in > > the system. > > > > What this says is that some of the database statements are being executed > > not in an explicit transaction. If you use transactions it is required > > that ALL statements are done this way, otherwise the DB will add put > > default transactions around the unbounded statements and they will not be > > in the right mode (in this case IMMEDIATE). Normally this will not be a > > problem, and quite possibly other DBs will get away with this, but SqLite > > has a non-standard way of doing the locking (this IMMEDIATE mode in the > > BEGIN TRANSACTION) and my guess is that this is the root of the problem. > > > > When I have time I will look at the specific SQL statements that are > > failing and see if that finds some of the missing beginTransaction calls. > > > > David
Given what upstream says about the SQLite backend's limitations: https://techbase.kde.org/Projects/PIM/Akonadi#Why_not_use_sqlite.3F https://techbase.kde.org/Projects/PIM/Akonadi/Database#Sqlite I'm not sure this kind of behavior isn't to be expected. I think it might be appropriate to add a release note that the SQLite backend should only be used on mobile systems and then lower the severity. Scott K
signature.asc
Description: This is a digitally signed message part.
