apr-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Aaron Bannert <aa...@clove.org>
Subject Re: [PATCH] new lock APIs, rwlocks, condition variables
Date Tue, 04 Sep 2001 06:42:54 GMT
On Mon, Sep 03, 2001 at 11:35:52PM -0700, Justin Erenkrantz wrote:
> On Mon, Sep 03, 2001 at 10:11:18PM -0700, Justin Erenkrantz wrote:
> > I'm chatting with Aaron to see what's up with this and if I can
> > reproduce his results.  I'm scratching my head at this.
> 
> (Just to keep the list in sync with what I've found...)
> 
> In my tests, the performance is essentially identical (Aaron sent
> me the httpd component of the patch).
> 
> However, --without-threads is busted with this patch.  -- justin

I will be fixing --without-threads tomorrow. Thanks for bringing this
to my attention.

What is typically done for functions like these in APR, Should I:

a) put #if APR_HAS_THREADS around the apr_thread_mutex_*() function,
   prototypes; essentially causing all code that wishes to use these functions
   to either fail at compile-time or have to put their own conditionals in?

or
b) have those functions that cannot function without APR_HAS_THREADS simply
   return APR_ENOTIMPL?

(i'm leaning for the (b) option, but that's just me)


Also, with Justin's help, I realized that the mod_include improvements
from the last week didn't make it into my control-test tree, and those
were the differences I was seeing in the speeds. I'm now registering a
[very] slight improvement with httpd-2.0 using my new code. Something
on the order of [censored: below margin of error]. ;)

-aaron

Mime
View raw message