>   Admittedly, this state of affairs has improved since the Borland C++
>   1.0 days, with the introduction of an internal layer of functions with
>   secret names on top of which both the Standard C Library functions and
>   the POSIX functions sit (a useless layer -- Borland should have made
>   everything just call the OS/2 API directly), but there are still some
>   problem areas, and it is still difficult to correct the POSIX
>   implementation.

I haven't used the OS/2 version of borland, I was reminiscing back to my
dos days...  I use EMX at home and CSET at work.  (I like the fact that
under EMX, the libraries you specify with -l are linked with before the
standard libraries, so you can replace anything you darn well please.)

What confuses me is that the C++ "standard" spent so much time
worrying about obscure new classes and didn't bother with things like
making "strcpy" and "memset" inline (or at least a
#define that makes selected function inline.)  Or a default value of NULL
for the argument to time()...

That sort of thing.

Rob
 
--- Xblat