On Mon, Aug 13, 2012 at 3:31 AM, Lauri Kasanen <[email protected]> wrote:
> On Sun, 12 Aug 2012 18:18:31 +0200
> Sonny Karlsson <[email protected]> wrote:
>
>> Current master causes build warnings when configured with --trace. It
>> may be possible to solve this by changing the order of include
>> statements or adding an extern declaration somewhere, but I have not
>> found a way to do so yet.
>>
>> I don't think it's very nice to have lots of implicit declaration
>> warnings, so I suggest reverting this patch. Some quick benchmarking
>> suggests there is a 1% drop in req/sec from doing so.
>>
>> > ./include/mk_iov.h: In function ‘mk_iov_add_entry’:
>> > ./include/mk_iov.h:114:5: warning: implicit declaration of function
>> > ‘mk_print’ [-Wimplicit-function-declaration]
>> > In file included from mk_plugin.c:30:0:
>> > ./include/mk_utils.h: At top level:
>> > ./include/mk_utils.h:83:6: warning: conflicting types for ‘mk_print’
>> > [enabled by default]
>> > In file included from ./include/mk_plugin.h:29:0,
>> > from ./include/mk_utils.h:57,
>> > from mk_plugin.c:30:
>> > ./include/mk_iov.h:114:5: note: previous implicit declaration of
>> > ‘mk_print’ was here
>
> I'll take a look at building with trace today. But not inlining this function
> is bad, as it's called extremely often (~9 times per request in my test - for
> 200k requests I had 1.8M calls to this).
>
> So NAK this patch from me.
Hi,
thanks for catching this.
Revert the optimization changes is not desired, as temporal solution i
have placed the mk_unlikely() conditional into the static function, it
uses printf(3) instead of mk_print():
http://goo.gl/ofvTF
cheers,
>
> - Lauri
> _______________________________________________
> Monkey mailing list
> [email protected]
> http://lists.monkey-project.com/listinfo/monkey
--
Eduardo Silva
http://edsiper.linuxchile.cl
http://www.monkey-project.com
_______________________________________________
Monkey mailing list
[email protected]
http://lists.monkey-project.com/listinfo/monkey
