Hi Alex,

I didn't check the table schemas for usrloc but I'm sure there may be other cases where the affected_rows function has been 'misunderstood'. In the code I picked this bug up (ims_pcscf_usrloc), I did exactly that, change the schema. Just wanted to discuss in case it was decided to change the connect flags to mitigate any future probs.

Also, if you merely change the the schema, some code would think the update had "failed" and do some other adverse failure code so not sure that would be an ideal final fix...

Cheers

Jason

On Fri, Aug 8, 2014 at 4:06 PM, Alex Hermann <alex@speakup.nl> wrote:

On Friday 08 August 2014, Jason Penton wrote:
> I have noticed that in some instances if you update a row in mysql via the
> mysql_db module and the actual row data does not change - affected_rows
> will return 0. This is the default behaviour for the mysql API as per -
> http://dev.mysql.com/doc/refman/5.0/en/mysql-real-connect.html
>
> There is a flag (CLIENT_FOUND_ROWS) that can be used in the
> mysql_real_connect function that will cause affected_rows to return the
> number of rows that were "matched" - ie in the WHERE clause, as opposed to
> whether or not any data was changed.
>
> If we don't it could be a problem for modules like usrloc where an update
> is done and if no row are "affected" and new row is added which would cause
> a duplicate.

If that happens, the table definition is wrong. It should have (a) unique
key(s) to prevent double records. We'd better fix that.

--
Greetings,

Alex Hermann

_______________________________________________
sr-dev mailing list
sr-dev@lists.sip-router.org
http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev