httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Brian Pane <brian.p...@cnet.com>
Subject Re: [PATCH]: Counting I/O with FLUSH
Date Mon, 21 Oct 2002 18:11:53 GMT
Bojan Smojver wrote:

>Quoting Brian Pane <brian.pane@cnet.com>:
>
>  
>
>>As for the MMN bump, IMHO it would be better to code the
>>patch in a way that doesn't require it (if I remember correctly,
>>you had an alternate solution involving a module-specific
>>structure instead of new fields in conn_rec).  That would
>>make it easier to get the improved bytes-sent accounting
>>into the hands of users who need it, without having to wait
>>until 2.1 or 2.2.
>>    
>>
>
>Yep, there was such a structure in mod_logio. However, given that the counting
>of output is now done in core_output_filter, such a structure would have to be
>introduced in the core, rather then in the mod_logio itself (i.e. core should
>not rely on mod_logio being present during compilation, dynamic linking or
>execution).
>
>I'll rework the patch to include the structure structure in core.c rather then
>one in mod_logio.c. Does that sound OK?
>

Sounds good to me

>Alternative is to let core.c know that there is such a thing as mod_logio.c for
>the purposes of fetching the structure. In other words, something like this
>would appear in core.c:
>
>--------------------------------------------------
>logio_config_t *cf = ap_get_module_config(c->conn_config,                      
>                       
>                                          &logio_module);
>--------------------------------------------------
>
>Which means that something like this would also have to appear in core.c:
>
>--------------------------------------------------
>extern module logio_module;
>--------------------------------------------------
>
>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.

Brian




Mime
View raw message