Re: [Firebird-devel] SCHAR, TEXT definition

[Fred P. Jr. <fps...@ya...>]

Mike Nordell wrote:
>>There is no real change in the way code works for most platforms, since
>>char is usually mapped to signed,
> 
> 
> "char" is usually a signed type, but it sure isn't "mapped" to anything but
> "char". No matter what platform you use, and what compiler switches you use:

I think Mike is correct. This is my understanding of how char works. As I recall, char is the only integral type like this. I *think* if you check the typeids of short, signed short, unsigned short, then you'll only see two distinct typeids.

Perhaps, a variation of Mike's code would help. In the code below I mixed the three flavors of char. [I used pointers to get around the compiler doing implicit conversions and masking the differences.] As you can see from the compiler errors, g++ (3.2) considers char*, signed char* or unsigned char* to be three distinct pointer types.

Finally, I checked the <limits> header file.  There are three distinct specializations of the numeric_limits template class: numeric_limits<char>, numeric_limits<signed char>, numeric_limits<unsigned char>.  So, no mapping appears to be taking place.

Hope this helps,
---Fred P.

----------------------------------
int main()
{
     char *cp = "1";
     signed char *scp = cp;
     unsigned char *ucp = cp;
     cp = scp;
     cp = ucp;
     scp = ucp;
     ucp = scp;
}
-----------------------------------
g++ -g char.cpp
char.cpp: In function `int main()':
char.cpp:4: invalid conversion from `char*' to `signed char*'
char.cpp:5: invalid conversion from `char*' to `unsigned char*'
char.cpp:6: invalid conversion from `signed char*' to `char*'
char.cpp:7: invalid conversion from `unsigned char*' to `char*'
char.cpp:8: invalid conversion from `unsigned char*' to `signed char*'
char.cpp:9: invalid conversion from `signed char*' to `unsigned char*'