Quoting Brian Pane <[EMAIL PROTECTED]>:
> >The core would also have to know about logio_config_t, which doesn't also
> >doesn't make sense. I would rather have it the other way around.
> >
>
> I agree--better to add a struct in the core and make mod_logio
> depend on that than to require the core to know about mod_logio.
OOPS, I think we'll have to come up with a compromise solution here. The problem
is that the core is already stuffing its own configuration (a socket) into
c->conn_config and I doubt we can put two configuration records in it by the
same module.
Which means I have to do this:
--------------------------------------------------------------
/* Structure for passing on the socket and logging of input/output bytes per
* connection and eventually per request. The logging stuff is used in both
* core.c and mod_logio.c */
typedef struct core_conn_conf {
apr_sock_t csd;
apr_off_t bytes_in;
apr_off_t bytes_out;
} core_conn_conf;
--------------------------------------------------------------
I'll fix all code in Apache that uses the socket directly after this kind of call:
--------------------------------------------------------------
ap_get_module_config(something->conn_config, &core_module);
--------------------------------------------------------------
to get the socket from the structure instead:
--------------------------------------------------------------
ccc->esd
--------------------------------------------------------------
At least two problems with this:
- does this require an MMN bump by itself
- is this likely to break something in the 3rd party module world that expects
core conn_config to be a socket
As usual, any input appreciated...
Bojan
