Re: [Tinycc-devel] Re: global 64-bit variables initialization

[grischka]

Masha Rabinovich wrote:

> About the tasks.. I think, bug-fixing is important (at least for me:) 
> but there are a few low-priority big tasks, like emiting #line in 
> preprocessor output or producing .map-file and debug-info the debuggers 
> can use (for example Microsoft's PDB-file)

I think I might have a patch for #line somewhere ;)

PDB would be great but is work.  GDB info should work with line-numbers
and function-names (which is "tiny" but better than nothing)

Of course if someone feels that TCC should look nicer than it really
does, there is also the possibility to rewrite parts.

One candidate for a rewrite is the macro expansion in the preprocessor.
It has a bug and I have a feeling that fixing it requires a different
logical layout.  If someone dares, it is "only" a few functions.
(macro_arg_subst, macro_subst_tok, macro_twosharps, macro_subst, next)

The bad case:

/* pp-bug.c */
#define c x.c
#define d c
#define f(e) e
f(d)

$ gcc -E pp-bug.c
x.c
$ tcc -E pp-bug.c
x . x . c

> Also would be nice to have some C++ support, but it could be very 
> difficult, I lost the hope to have a superfast C++ compiler.

I once have made some experiments with the historical "CFRONT AT&T C++
to C Translators", with TCC used for the PP and the C to OBJ part.

It works quite well, with both cfront 1.0 (1986) and cfront 3.0.3 (1997).
If someone is interested I can post more information.  Note however that
even the latter version is not about today's C++ (although close already).

--- grischka