sqlite3 support is not thread safe

[GoogleCodeExporter]

I am using python3forandroid. It turns out the this version comes with a 
sqlite3 support which has thread disabled. As a result, I frequently corrupt my 
databases and get the error 'the database image is malformed'. I would 
appreciate it very much if sqlite3 support threads.

Original issue reported on code.google.com by jerryfle...@gmail.com on 27 Jun 2012 at 5:17

[GoogleCodeExporter]

The release notes on that version of sqlite3 recommended against using the 
threaded version. 
It should be relatively easy to recompile with the threaded version, but I'd 
like to know more about the options.
In other words, I'll look at it if you research the pros and cons.

Original comment by rjmatthews62 on 27 Jun 2012 at 6:05

[GoogleCodeExporter]

Maybe I should put it more clear. I was running my script while another process 
is changing (or 'prepare' to do so) the database at the same time. I was 
wondering if thread safety can lock the table for example. The inability to 
modify the other processes qualifies this feature request, I guess, even I am 
aware that threads are evil. Anyway, having this feature doesn't enforce you to 
use it. You still have an option to disable it at connection time if you really 
hate it.

Original comment by jerryfle...@gmail.com on 27 Jun 2012 at 7:41

[GoogleCodeExporter]

hmm... db file locking to support concurrency that you require should 
theoretically be built into SQLITE3 anyway - adding threading support isn't 
likely to make this work. The SQLITE page on this topic further notes that a 
number of OS/FileSystems have issues in this with few/no work-arounds - 
http://www.sqlite.org/lockingv3.html#how_to_corrupt - so it could be an OS 
limitation.

Original comment by srinathd...@gmail.com on 27 Jun 2012 at 9:33

[GoogleCodeExporter]

You may also want to take a look @ the discussion here - 
http://www.sqlite.org/faq.html#q5 - note the recommendation of using a client 
server architecture for the database in a situation where file locking seems to 
be broken & a lot of concurrency is needed.

Original comment by srinathd...@gmail.com on 27 Jun 2012 at 9:52