Hello
This might seem the cause of the recent buggy behaviour of DBD::mysql:
bye,
-christian-
On 2006-08-25 Niko Tyni wrote:
> On Thu, Aug 24, 2006 at 02:02:43AM +0200, Christian Hammers wrote:
>
> > Some users of the Debian packages I create recently reported that after the
> > upgrade from MySQL 5.0.22 to 5.0.24 Perl scripts that try to explicitly set
> > $dbh->{AutoCommit} = 0;
> > die with the error message:
> > "Transactions not supported by database"
>
> Hi,
>
> this bug (#384047) is also at least #384221 (libdbd-mysql-perl) and
> #383700 (request-tracker3.4).
>
> I believe the problem is an incompatible ABI change in
> libmysqlclient.15.so with version 5.0.24.
>
> This addition in struct st_mysql_options:
>
> --- mysql-dfsg-5.0-5.0.22/include/mysql.h 2006-05-25 11:56:42.000000000
> +0300
> +++ mysql-dfsg-5.0-5.0.24/include/mysql.h 2006-07-27 21:52:03.000000000
> +0300
> @@ -164,6 +164,7 @@
> char *ssl_ca; /* PEM CA file */
> char *ssl_capath; /* PEM directory of CA-s? */
> char *ssl_cipher; /* cipher to use */
> + my_bool ssl_verify_server_cert; /* if to verify server cert */
> char *shared_memory_base_name;
> unsigned long max_allowed_packet;
> my_bool use_ssl; /* if to use SSL or not */
>
> causes the size of st_mysql_options to grow. This also affects the MYSQL
> struct, since it contains st_mysql_options.
>
> Now, libdbd-mysql-perl has this in dbdimp.h:
>
> struct imp_dbh_st {
> dbih_dbc_t com; /* MUST be first element in structure */
>
> MYSQL mysql;
> int has_transactions; /* boolean indicating support for
> * transactions, currently always
> * TRUE for MySQL and always FALSE
> * for mSQL.
> */
> [...]
>
> so the grown MYSQL struct overlaps with has_transactions.
>
> This hits libdbd-mysql-perl 3.006-1, which is currently compiled
> against libmysqlclient15off-dev 5.0.22-x (with x depending on
> the architecture). When run with libmysqlclient15off 5.0.24-1,
> has_transactions gets reset to 0 and the result is 'Transactions not
> supported by database'. Recompiling libdbd-mysql-perl helps, as the API
> has not changed in an incompatible way.
>
> In my understanding, the right fix is to bump the soname or to revert
> the struct st_mysql_options change.
>
> Cheers,
> --
> Niko Tyni [EMAIL PROTECTED]
>
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
