Angus Leeming <[EMAIL PROTECTED]> writes:
>> | {
>> | +#if defined (WIN32)
>> | ifstream ifs(from.c_str(), ios::binary);
>> | +#else
>> | + // A bug in the gcc 2.95 implementation of ifstream
>> | + // means that we can't use ios::binary here (even though
>> | + // it should have no effect on unix machines).
>> | + ifstream ifs(from.c_str());
>> | +#endif
>>
>> I don't like this ... If at all... only the failing compiler/stdlib
>> should have special code.
>
| Sheesh! With all due respect, why do you care? The flag should have
| absolutely no effect on a unix machine, so the two pieces are exactly
| equivalent to a compliant compiler/lib.
This is not the point. The point is that "ios::binary" is standard
conforming. So we should be able to use this without any special
defines.
| Nonetheless, to show willing:
>
| #if defined (__GNUC__) && defined (__GNUC_MINOR__) && \
| !defined(_STLPORT_VERSION) && \
| __GNUC__ == 2 && __GNUC_MINOR__ <= 95
>
| Do you really want that?
HAVE_WORKING_IOS_BINARY_FLAG
Besides do we really support 2.95 anymore? (and its stdlib?)
--
Lgb
