What I've seen is that windows makes a local copy of the memory mapped file. Perhaps in swap. If you watch your application using "ProcMon" you can see all the disk IO. That's how I noticed the copy.
For my usage, only one process opens the DB so, maybe it's a non-issue. C Tuesday, November 8, 2011, 10:05:17 PM, you wrote: PI> On Mon, Nov 7, 2011 at 5:48 PM, Jean-Christophe Deschamps PI> <j...@antichoc.net> wrote: >> Pavel, >> >>> This is not a local file. Even if you use your hostname as netname >>> file is still retrieved through network stack. And I guess SAMBA >>> doesn't work well with memory mapped files (in addition to all >>> problems with locking). >> >> True but why did I get no error? Since MMF don't cope with network it would >> be good to get a no-no somewhere. I really don't know which layer should >> bark however. PI> I can't answer this question. Maybe it's something like all operations PI> are successful but memory mapping is not actually shared between PI> processes. So every process sees its own copy and assumes that no PI> other process works with database, thus database can be corrupted in PI> many possible ways. PI> Pavel PI> _______________________________________________ PI> sqlite-users mailing list PI> sqlite-users@sqlite.org PI> http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users -- Best regards, Teg mailto:t...@djii.com _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
