On Tue, Feb 21, 2012 at 10:53:33AM +0100, Florian Weimer wrote: > > On Mon, Feb 20, 2012 at 10:03:51AM +0100, Florian Weimer wrote: > >> Have you tried setting "PRAGMA journal_mode = WAL" on the SQLite > >> database, using the sqlite3 command line tool? > > > > Nope. I didn't know it existed. If it's needed for reasonable > > performance, it should be set by default upon installation. > > It prevents sharing the home directory over NFS on multiple machines > at the same time.
So you're saying liferea (or better, libsqlite) should make a statfs call and enable or disable it on the fly. NFS support is nice, but it isn't a common use case and doing bad things for performance for the common case in order to support NFS is not very nice. > > I tried using this and patched the source code. It was ineffective. > > Then fsync performance isn't the issue. Unless the liferea database > is really large, a couple of writes shouldn't have the impact you > observe. So it's probably something else. I have approximately 300 feeds. I imported the blogrolls of Planet Debian and some other planets some years back. I also have some feeds (such as the four or five from the Houston Chronicle) that explicitly request updates every 6 minutes. The problem may or may not be fsync-related. But on my old laptop (the one I had when I reported this bug) with a 5400 RPM disk, liferea made other disk-using processes dog slow. As I said, it had negative effects on dpkg and also on the load average far beyond what is reasonable. -- brian m. carlson / brian with sandals: Houston, Texas, US +1 832 623 2791 | http://www.crustytoothpaste.net/~bmc | My opinion only OpenPGP: RSA v4 4096b: 88AC E9B2 9196 305B A994 7552 F1BA 225C 0223 B187
signature.asc
Description: Digital signature
