-include -g broken with current CVS (testcase for PR 771)

[Zack Weinberg]

On Mon, Nov 13, 2000 at 07:43:16AM +0000, Neil Booth wrote:
> Zack,
> 
> Is there a reason we cannot just move 
> 
>   if (! cpp_start_read (&parse_in, filename))
>     exit (FATAL_EXIT_CODE);	/* cpplib has emitted an error.  */
> 
> from init_c_lex to a new function called init_c_lex_delayed or
> init_parse_delayed that goes something like
> 
> void
> init_c_lex_delayed ()
> {
> #if USE_CPPLIB
>   /* We need to delay this call until after setting up debug
>      output, because -included files are processed immediately.  */
>   if (! cpp_start_read (&parse_in, filename))
>     exit (FATAL_EXIT_CODE);	/* cpplib has emitted an error.  */
> #endif
> }
> 
> and unconditionally call this before yyparse?  This would avoid
> callbacks.  I hope nothing wants to use parse_in before yyparse.

It'd require changing all front ends, which may not be wise when we're
closing in on 3.0.  I have no intrinsic objection, though.

zw