[PATCH] fortran/25106 -- Statement labels are nonzero.

[FX Coudert]

> New patch.  I checked F66, F77, F95, and F2003 standard.  A statement
> label must be nonzero.  Bootstrap and regtested on i386-*-freebsd.

As we discussed on IRC, I think the match routine needs to have exit so 
that the parser could recover in some cases (although I can think of no 
case involving this, just to be on the safe side). Moreover, no function 
in match.c currently calls gfc_error_now, so I think it's better to keep 
it that way.

Here is another proposition for a patch, based on yours. Passes 
regtesting on i686-linux. What do you think of it?

FX