----- Original Message ----- From: "Cliff Woolley" <[EMAIL PROTECTED]> To: "Bill Stoddard" <[EMAIL PROTECTED]> Cc: <[EMAIL PROTECTED]> Sent: Monday, March 18, 2002 1:41 AM Subject: Re: cvs commit: httpd-2.0/modules/experimental mod_cache.h mod_mem_cache.c
> On 18 Mar 2002 [EMAIL PROTECTED] wrote: > > > jwoolley 02/03/17 22:37:33 > > > > Modified: modules/experimental mod_cache.h mod_mem_cache.c > > Log: > > Get mod_mem_cache to compile cleanly again on linux, where apr_atomic_t > > is a struct. > > Hey Bill... > > Do me a favor and look over this (ie, please make sure it still works > on Windows :). It's a bit ugly, but it was necessary to get it to compile > cleanly. There were a few cases where I couldn't tell how it was that the > !USE_ATOMICS case was actually working right... how long has it been since > somebody tried it with that turned off? Just curious. But anyway, why > do we need a special !USE_ATOMICS case? We have a default implementation > of apr_atomic_* that uses locks to do its thing... are we just trying to > avoid the extra locking overhead? > > PS: the (void) cast on the two calls to apr_atomic_dec() was needed to > get rid of a compiler warning saying that a value computed was never used. > > Thanks, > --Cliff mod_mem_cache is experimental. So are the apr_atomic_* calls. Feel free to remove the !USE_ATOMICS code paths when the apr_atomic functions are reasonably stable. Will review a bit later today. Bill
