Re: gcc 3.0.1 preprocessor bug

[Andrew Pinski <pinskia at physics dot uc dot edu>]

If you put a space before the dot in m(a)
so it will become:
#define m(a) x_##a##0_x .x
This is not a work around; this is the solution.

The error message is correct pasting the two will not give a valid preprocessing token.


Thanks,
Andrew Pinski
Student of Computer Science
University of Cincinnati

On Monday, August 20, 2001, at 11:20 PM, Andy Tai wrote:

> Hi, this seems a bug in the preprocessor of gcc 3.0.1
> but I am not sure so I post here first.
>
> sample code:
> ---t.cc---
>
> struct s
> {
> int x;
> };
>
> #define m(a)  x_##a##0_x.x
>
>
> main()
> {
> struct s x_10_x;
> m(1) = 2;
>
> }
>
> ---end of t.cc---
>
> with gcc 2.95.3 on GNU/Linux (Red Hat 7.1), no error.
>
> with gcc 3.0.1 on the same machine, gcc says
>
> % gcc t.cc
>
> t.cc:12:4: warning: pasting "x_1" and "0_x.x" does not
> give a valid preprocessing token
> t.cc: In function `int main()':
> t.cc:12: underscore in number
> t.cc:12: `x_1' undeclared (first use this function)
> t.cc:12: (Each undeclared identifier is reported only
> once for each function it    appears in.)
> t.cc:12: parse error before numeric constant
>
> % gcc -v
> Reading specs from
> /share/software/gcc-3/installed/lib/gcc-lib/i686-pc-linux-gnu/3.0.1/specs
> Configured with: ../gcc-3.0.1/configure
> --prefix=/share/software/gcc-3/installed
> --enable-threads
> Thread model: posix
> gcc version 3.0.1
>
> Thanks for any work arounds on this.