apr-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jeff Trawick <traw...@attglobal.net>
Subject Re: avoid -Wall warning in apr_md5.c
Date Thu, 29 May 2003 21:38:47 GMT
André Malo wrote:
> Just got a report via PM about a gcc warning during 2.0.46 compile:
> 
> System:
> gcc: 2.95.3
> Linux 2.4.18 Conectiva 8
> 
> configured with worker MPM.
> 
> apr_md5.c:688: warning: `crypt_mutex_lock' defined but not used
> apr_md5.c:693: warning: `crypt_mutex_unlock' defined but not used
> 
> The attached patch should avoid these, shouldn't it? (untested).

seems very likely, but I haven't tested it either :)

I think some more serious reimplementation is appropriate for that 
function anyway.

I think it would be cleaner if apr_initialize() set up =>1 thread 
mutexes for its use and apr-util's use, and export those via a global 
array and mutex count, then apr_password_validate(), and any other code 
with a similar problem, would use that.

#if APR_HAS_THREADS
struct {
     apr_thread_mutex_t **mutexes;
     int mutex_count;
} apr_private_mutexes;
#endif

count would be guaranteed >= 1

then apr_password_validate() wouldn't have to have its own mutex 
allocation handling; it would just use one of the available internal APR 
mutexes

#if APR_HAS_THREADS && !defined(APU_CRYPT_THREADSAFE)
   apr_thread_mutex_lock(apr_private_mutexes.mutexes[0]);
   crypt();
   apr_thread_mutex_unlock(apr_private_mutexes.mutexes[0]);
#else
   crypt();
#endif


Mime
View raw message