[Live-devel] 64 bit portability

[Ross Finlayson]

>- Are there any plans to get rid of there warnings to make Visual 
>Studio compilers happy?

No, because most of those warnings are silly and/or bogus, and the 
few warnings that *are* legitimate turn out not to be a problem in 
our code.

Specifically:

- The "conversion from 'SOCKET' to int" warnings are bogus, because 
the functions "socket()", "fcntl()", and "accept()" have always 
returned "int" - so if Windoze now thinks that they should return 
some other data type, then it's just plain broken, IMHO.

- The "conversion from 'size_t'" warnings are a bit silly, because 
functions like "strnlen()" etc. can never reasonably be expected to 
return sizes >= 2^32 (our buffers will never be that large).

- The "conversion to pointer types of 'greater size'" warnings are 
silly, because these can never cause errors.  In our code, these 
warnings happen only because we use "char*" (i.e., a pointer type) as 
the type of keys in our hash tables, but the actual keys that we use 
in some hash table instances will often be other data types (e.g., 
unsigned) that might (on a 64-bit machine) be smaller (but never 
larger).

- The "pointer truncation" warnings are legitimate, but in our code 
they're OK, because (in all but one[*] case) they occur only when we 
retrieve a key from a hash table, and we know for sure that the 
actual key type is some type other than a pointer.

[*] The one other place where we do a 'pointer truncation' is when we 
use a pointer value to generate a random hash table index.  In this 
case it's OK to use the lower 32 bits only.
-- 

Ross Finlayson
Live Networks, Inc.
http://www.live555.com/