I'm getting a little closer to pointing the finger at sqlite. I rebuilt my application using an archived binary from march 2009, and it works in the cases where a modern binary fails.
It appears that some change in sqlite locking strategy between 2009 and now is not completely compatible with osx 10.4.11 At 01:46 PM 3/2/2011, Dave Dyer wrote: >The configuration I'm interested in has: > > Database resident on a windows 7 file system > sqlite binary based on sqlite 3.7.3 > > If the host is OSX 10.6, it works. > If the host is OSX 10.4, it fails. > >I swear it used to work in 10.4 too, but lacking a time >machine, it's hard to prove. I'm certain it worked a year ago with >the then-current version of OSX 10.4 and sqlite 3.6.11 > >using a debugger, I traced the source of the problem to afpSetLock, which >is called with identical context on both OSX 10.4 and 10.6, but fails >in 10.4. It seems odd that afpSetLock should be used for this >remotely mounted file system, but it is the same for both the successful >and unsuccessful cases. > >I've built a version with SQLITE_ENABLE_LOCKING_STYLE=0, which eliminates >afpSetLock, and in that case it fails in "unixLock" on both operating >systems. > >Note also that the same binary and an identical database work fine >if the database is on a local file system. > >Is there any change in sqlite locking strategy that can explain this? Or >alternatively, am I just being screwed by some Apple bug. I'm sure that >Apple has been messing with the remote file system support. > >_______________________________________________ >sqlite-users mailing list >sqlite-users@sqlite.org >http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
