Re: svn commit: r33520 - requiring SQLite

On Tue, Oct 7, 2008 at 2:31 PM, Justin Erenkrantz <justin_at_erenkrantz.com> wrote:
> On Tue, Oct 7, 2008 at 2:20 PM, Greg Stein <gstein_at_gmail.com> wrote:
>> There is some #if'd code in init_sqlite() that is incomplete. Are
>> those two functions in 3.4?
>
> Nope. We should be able to protect them with another
> SQLITE_VERSION_AT_LEAST(3,5,0) check. Hmm, sqlite3_threadsafe() isn't
> defined in 3.4.x, but gcc isn't emitting a warning - but, it too
> should probably be behind a 3.5.0+ check.

There used to be a guarded check for that, but I tossed it when I
assumed 3.5.x minimum.

> We can also do run-time version checking with
> sqlite3_libversion_number() to ensure we don't downgrade from what we
> compile against.

Right. My WC has a comment in sqlite.c to do just that. Specifically,
I'm worried about apr-util loading an old sqlite into the process
space, regardless of what we compile/link against. Thus, the runtime
check is mandatory.

> I don't know what guarantees SQLite3 has with
> respect to ABI compatibility though (if any)... -- justin

Functions can come and go, but I think the functions' signature and
semantics are rigid.

Hmm. On the config stuff... I think here are corresponding pragmas
(just less easy to use).

Cheers,
-g

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe_at_subversion.tigris.org
For additional commands, e-mail: dev-help_at_subversion.tigris.org
Received on 2008-10-07 23:44:32 CEST