Coding question

["D.J. Stachniak" <dj...@yahoo.com>]

Hi all,

 

I was reading through some of the RAM cache code and stumbled across this:

 

http://svn.apache.org/repos/asf/incubator/trafficserver/traffic/branches/dev
/iocore/cache/RamCache.cc

 

int

RamCache::get_lock(INK_MD5 * key, Ptr<IOBufferData> *ret_data, EThread * t,
inku32 auxkey1, inku32 auxkey2)

{

  inku32 k = key->word(3);

  int pp = k % n_partitions;

  RamCachePartition *p = &partition[pp];

  (void) p;   // ???

  CACHE_TRY_LOCK(l, p->lock, t);

 

Or something similar:

 

int
RamCache::put(INK_MD5 * key, IOBufferData * data, EThread * t, inku32
auxkey1, inku32 auxkey2)
{
 
  /* equivalent to ram cache disabled */
  if (partition_size == 0)
    return 0;
 
  ProxyMutex *mutex = t->mutex;
  (void) mutex;  // ???
  inku32 k = key->word(3);

 

I'm wondering what's its (the seemingly no-op void casts) purpose?  Is it a
long, lost holdover from some previous compiler version which required a
statement like that to keep it quiet for some warning?  I've seen it in
several different files.

 

Just curious...didn't know if this was some obscure trick I was unaware of -
other than quieting a compiler to say a variable's unreferenced...

 

Thanks,

D.J. Stachniak

 

Re: Coding question

[John Plevyak <jp...@acm.org>]

Yes.  Those cases were there to eliminate
the unused variable warnings if (for example)
the mutex was unused or the CACHE_TRY_LOCK
became a noop.

A lot of cases where variables
might or might not have been used have been
cleaned up or NOWARN_UNUSED(x) substituated
(which results in (void)x)) or NOWARN_UNUSED_RETURN
which results in if(x) {} which isn't 'else' safe).

Interestingly we also have UNUSED and INK_UNUSED
which use the __attribute(unused) when available, but
don't do (void)x when it isn't.

The upshot is that all this should probably be
cleaned up.

We should probably eliminate UNUSED and INK_UNUSED
and change NOWARN_UNUSED to use the attribute
if available or and change NOWARN_UNUSED_RETURN to
do {} while (x) for safety.

something for jira?

john

On 1/14/2010 8:35 PM, D.J. Stachniak wrote:
> Hi all,
> 
>  
> 
> I was reading through some of the RAM cache code and stumbled across this:
> 
>  
> 
> http://svn.apache.org/repos/asf/incubator/trafficserver/traffic/branches/dev
> /iocore/cache/RamCache.cc
> 
>  
> 
> int
> 
> RamCache::get_lock(INK_MD5 * key, Ptr<IOBufferData> *ret_data, EThread * t,
> inku32 auxkey1, inku32 auxkey2)
> 
> {
> 
>   inku32 k = key->word(3);
> 
>   int pp = k % n_partitions;
> 
>   RamCachePartition *p = &partition[pp];
> 
>   (void) p;   // ???
> 
>   CACHE_TRY_LOCK(l, p->lock, t);
> 
>  
> 
> Or something similar:
> 
>  
> 
> int
> RamCache::put(INK_MD5 * key, IOBufferData * data, EThread * t, inku32
> auxkey1, inku32 auxkey2)
> {
>  
>   /* equivalent to ram cache disabled */
>   if (partition_size == 0)
>     return 0;
>  
>   ProxyMutex *mutex = t->mutex;
>   (void) mutex;  // ???
>   inku32 k = key->word(3);
> 
>  
> 
> I'm wondering what's its (the seemingly no-op void casts) purpose?  Is it a
> long, lost holdover from some previous compiler version which required a
> statement like that to keep it quiet for some warning?  I've seen it in
> several different files.
> 
>  
> 
> Just curious...didn't know if this was some obscure trick I was unaware of -
> other than quieting a compiler to say a variable's unreferenced...
> 
>  
> 
> Thanks,
> 
> D.J. Stachniak
> 
>  
> 
>