Re: Arithmetic Shift

[Andrew W. Nosenko]

On Sun, Dec 12, 2010 at 03:05, Andrew W. Nosenko
<address@hidden> wrote:
> On Sat, Dec 11, 2010 at 22:09, Paul Eggert <address@hidden> wrote:
>> On 12/11/2010 06:16 AM, Andrew W. Nosenko wrote:
>>> You mismatch the preprocessor function (find the LONG_MAX macro and
>>> replace it by it's definition found somewhere in header) with the
>>> function of compiler (inject proper search path to the preprocessor if
>>> compiler redefines some system's headers).
>>>
>>>> >
>>>> > The situation with -1 >> 1 is similar.
>>>> >
>>> No, because replacement of macro by it's value is essentially
>>> search-and-replace, while evaluation of arithmetics expressions is
>>> essentially code interpretation.
>>
>> Sorry, I don't understand your point.  It seems to be
>> claiming that the preprocessor doesn't do arithmetic
>
> Of course it does :-)
>
>> (a claim that is obviously incorrect), but your earlier
>> comments suggest that you understand things well enough
>> not to be making claims like that.
>
> I'm arguing that in case of LONG_MAX example there no math at the
> preprocessor side.  Plain string substitution.  Any evaluation, if
> need, made latter by compiler.  Nothing bad will happen in this
> example if cpp doesn't match cc.  At least while proper search path
> passed for find limits.h (and therefore LONG_MAX define) that
> corresponds the target architecture.
>
> From another hand, the #if -1>>1==-1 explicitly requires to evaluate
> this expression by the cpp and there no other way for preprocessor for
> do it's job than do this evaluation.
>
> Therefore these cases (LONG_MAX vs #if -1>>1==-1) are not similar at
> all.  First depends fully on the proper search paths and doesn't
> depends on the preprocessor implementation at all.  Second absolutely
> doesn't depend on the paths but fully depends on the preprocessor
> implementation.
>
>> Anyway, I don't think it matters.  We seem to be agreeing
>> that autoconf isn't needed here.
>
> Sure!
>

Another possibility is that it's not you misunderstand or doesn't
understand my point, but it's I don't understand or misunderstand
your.

If interpret your mail not as

    "cpp is pretty tied to cc and any mismatch gives The Bad Thing,
for example you may give the wrong LONG_MAX",

but as

    "you should write (as much as possible) on the fault-tolerant
manner and doesn't depend (as much as possible) that your cpp+cc pair
is bug free and, for example, didn't lied to you about actual LONG_MAX
value"

then all pieces of puzzle hits own place and all becomes much clearer :-)

-- 
Andrew W. Nosenko <address@hidden>