Rob, (Nick)
I have been looking a little more at the issue of returning
meaningful error messages back to apache - i didn't like my previous
mail's hijacking of apache as a solution - though it should provide
an idea of what is going on.
So now the implementation looks like. Rob, sorry - this should be a
better way of doing exactly what my previous mail suggested - (caveat
- I have built these, but not run them - the production machine is
live)!
However, we now have kept httpd separate from apr and so we don't
need to mess with the includes and so on.
In srclib/apr-util/dbd/apr_dbd_mysql.c - rewrite
dbd_mysql_check_conn, around line 740
--------------
static apr_status_t dbd_mysql_check_conn(apr_pool_t *pool, apr_dbd_t
*handle) {
unsigned int mysql_err_num = mysql_ping(handle->conn);
if (mysql_err_num == 0) {
return APR_SUCCESS;
} else {
return APR_OS_START_USERERR + mysql_err_num; /* we could
also remove the CR_MIN_ERROR offset... */
}
}
--------------
Also in srclib/apr-util/dbd/apr_dbd_mysql.c - rewrite
dbd_mysql_error, around line 240
--------------
static const char *dbd_mysql_error(apr_dbd_t *sql, int n) {
const char *mysql_error_msg_string = mysql_error(sql->conn);
if (mysql_error_msg_string != NULL) {
return mysql_error_msg_string;
} else {
switch (n) {
case 0: return mysql_error_msg_string;
case CR_COMMANDS_OUT_OF_SYNC: return "mysql error
COMMANDS_OUT_OF_SYNC";
case CR_SERVER_GONE_ERROR: return "mysql error
SERVER_GONE_ERROR";
case CR_UNKNOWN_ERROR: return "mysql error UNKNOWN_ERROR ";
/* add a whole load of other mysql errors here see mysql
file errmsg.h */
default: return "unspecified mysql error";
}
}
}
--------------
So we need to change return value of the apr_dbd_open function in
srclib/apr-util/dbd/apr_dbd.c line 170. - The API is open regarding
errors (lucky us) so rather than blanking the error out with a GEN,
we can return rv
Change:
if ((rv != APR_SUCCESS) && (rv != APR_ENOTIMPL)) {
apr_dbd_close(driver, *handle);
return APR_EGENERAL;
}
To:
--------------
if ((rv != APR_SUCCESS) && (rv != APR_ENOTIMPL)) {
apr_dbd_close(driver, *handle);
return rv;
}
--------------
And finally we can change mod_dbd.c (modules/database/mod_dbd.c)
around line 288
--------------
if (rec->driver == NULL) { ap_log_perror(APLOG_MARK, APLOG_CRIT,
rv, rec->pool,"DBD: mod_dbd received a NULL when calling
apr_dbd_get_driver for %s",svr->name); }
rv = apr_dbd_open(rec->driver, rec->pool, svr->params, &rec-
>handle);
if (rec->driver == NULL) { ap_log_perror(APLOG_MARK,
APLOG_CRIT, rv, rec->pool,"DBD: mod_dbd received a NULL when calling
apr_dbd_get_driver for %s",svr->name); }
rv = apr_dbd_open(rec->driver, rec->pool, svr->params, &rec-
>handle);
switch (rv) {
case APR_SUCCESS: {
ap_log_perror(APLOG_MARK, APLOG_DEBUG, rv, rec->pool, "DBD: %
s driver is successfully opened and connected", svr->name);
} break;
case APR_ENOTIMPL: {
ap_log_perror(APLOG_MARK, APLOG_CRIT, rv, rec->pool, "DBD:
Can't connect to %s Not Implemented.", svr->name);
return APR_ENOTIMPL;
} break;
default: {
const char *errmsg = apr_dbd_error(rec->driver, rec->handle,
rv - APR_OS_START_USERERR);
ap_log_perror(APLOG_MARK, APLOG_CRIT, rv, rec->pool,"DBD:
Can't connect to %s - ", svr->name, (errmsg ? errmsg : "no reason
given"));
return APR_EGENERAL;
} break;
}
--------------
Nick, in modules/database/mod_dbd.c, just before we call
apr_dbd_open is there any reason why not to check that rec->driver is
not NULL ?
The previous error management code only evaluates return values, and
it seems possible to me that apr_dbd_get_driver can actually return a
NULL driver without returning an error code to match.
I am unsure as to whether or not the (rec->driver == NULL) test line
is redundant.
Another challenge is how to pass the error result from mysql_ping
(called in dbd_mysql_check_conn) over to dbd_mysql_error. I am not
impressed with merely discarding it!!
As the error is internal to mysql, it doesn't feel correct to be
returned by apr_dbd_open, but we do have APR_OS_START_USERERR so I
used that - is there a better way?
Now we are handling any messages correctly, through the apr_dbd_error
function, and we know that rec->driver is not null.
Of course the code above does a good deal more than merely deal with
connection issues - it demonstrates how to use mysql error reporting
from mod_dbd in a consistent manner. The significant change is making
use of the mysql error numbers and returning them in
APR_OS_START_USERERR space.
Hoping these mails are helping someone!
-b
---------------------------------------------------------------------
The official User-To-User support forum of the Apache HTTP Server Project.
See <URL:http://httpd.apache.org/userslist.html> for more info.
To unsubscribe, e-mail: [EMAIL PROTECTED]
" from the digest: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
