> There are two reasons why I don't want to replace it with XP-code:
> * I personally hate to have to remember which functions I should
> prefix with whatever to make them XP-functions, and which ones
> I can leave alone. That is why a configure-like solution is so
> much better: it links in these function under their own name,
> so you can still use strdup instead of having to remember
> to use UT_strdup.
I didn't mean strdup vs. UT_strdup, I meant
y = strdup( x );
vs.
malloc( strlen( x ) + 1 );
strcpy( y, x );
This solves the whole problem altogether.
> * More importantly, I really, really want to leave the code in
> the psiconv/psiconv directory exactly the same as in the
> lib/psiconv directory of the complete psiconv package. If we
> start to make changes between them now, keeping them both
> synchronised is going to be hell.
>
> The second reason is also why I am not very happy with your snprintf-
> patch to error.c; I will try to change it to something better.
Why does what I've done have to be Abi-specific? Why not fold it back
into the main psiconv lib? I can't imagine that the library being
compilable with the MSVC compiler is a bad thing...
-Tom
