[tz] proposed changes for Win32 and a improved mktime() algorithm

[Paul Eggert]

On 05/10/2017 08:35 AM, Paul.Koning at dell.com wrote:
> In GCC at least, you can suppress the warning with an extension: "int foo = foo;" which means "don't initialize but don't bug me".

Thanks, I didn't know about this extension. However, it has undefined 
behavior in standard C, so we can't use it in code intended to be 
portable everywhere. And it'd be error-prone even to use it in 
conditionally-compiled code. Suppose we added this macro:

#ifdef __GNUC__
# define INITIALIZED(x) x = x
#else
# define INITIALIZED(x) x
#endif

And suppose we remove -Winit-self from GCC_DEBUG_FLAGS, and replace 
existing code "int local; INITIALIZE(local);" with "int 
INITIALIZED(local);" to pacify GCC when we know that "local" cannot be 
used uninitialized but GCC doesn't deduce this. The problem with this 
approach is that if there is buggy code somewhere else that says "int 
local2 = local2;", GCC won't warn about it.


> because of the halting problem, it's not valid to say that a warning is a sign of an inferior compiler

It's a quality-of-implementation issue, and it's fair to say that some 
compilers are better than others in this area. It's true that because of 
the halting problem, no compiler can do a perfect job: for any compiler 
X that you can build, someone else can build a compiler Y that is better 
than X in this area. When I wrote "an inferior compiler", I meant "a 
compiler inferior to recent GCC", not "a compiler inferior to a perfect 
compiler".