Re: [Tracker] Performance improvement for tracker-0.6

[Jamie McCracken <jamie mccrack googlemail com>]

Good catch

I always used to do a valgrind before any release - hope you guys do the
same

jamie

On Fri, 2009-05-15 at 14:24 +0200, Philip Van Hoof wrote:
> On Thu, 2009-05-14 at 18:48 +0200, Carlos Garnacho wrote:
>
> > Here's an updated version, basically it moves table data handling to
> > libtracker-data (by introducing a TrackerDataUpdateMetadataContext
> > object), and adds support for initial insertions into the DB, instead of
> > doing 1 insertion and N updates.
>
> I was just reviewing 169fcacef91396ebe77e50a82241d5742ce53d39, and then
> I saw this:
>
> + tracker_db_interface_execute_query (iface, &error, sql, NULL);
> +
> + if (error) {
> + g_warning ("%s", error->message);
> + g_error_free (error);
> + }
> +} <- end of function
>
> Either the variable "sql" is const, but that isn't the case, or it's
> leaked. Right? Looking at it it's leaked in both cases. In case of
> INSERT it's leaked as a g_strdup_printf, in case of UPDATE it's leaked
> because the parameter FALSE is passed to g_string_free.
>
> And given that an awful lot of queries are executed by this function,
> this is a leak that will grow to quite big proportions (the size of all
> the stored data multiplied by length of the "UPDATE bla bla" pieces of
> the string).
>
> I suggest we add a g_free (sql) before a release is made with this ;)
>
> Feel free to also rewrite attached patch and apply.
>
> _______________________________________________
> tracker-list mailing list
> tracker-list gnome org
> http://mail.gnome.org/mailman/listinfo/tracker-list