gas: abort in new_logical_line_flags

[Jan Beulich]

On 17.02.2026 00:33, Alan Modra wrote:
> .linefile 0.
> 
> hits the "case 1 << 3" abort in new_logical_line_flags with line_number
> of -1.
> 
> Similar to commit b60f6a628897, this fixes the abort by extending the
> meaning of .linefile in user code to restore the physical line for
> ".linefile 0 .".
> 
> 	* input-scrub.c (new_logical_line_flags <case 1<<3>): Don't
> 	abort on negative line number, instead restore physical line
> 	for ".linefile 0".
> 
> diff --git a/gas/input-scrub.c b/gas/input-scrub.c
> index afb80ef2ce0..2b90632040b 100644
> --- a/gas/input-scrub.c
> +++ b/gas/input-scrub.c
> @@ -495,7 +495,12 @@ new_logical_line_flags (const char *fname, /* DON'T destroy it!  We point to it!
>        /* FIXME: we could check that include nesting is correct.  */
>        break;
>      case 1 << 3:
> -      if (line_number < 0 || fname != NULL)
> +      /* s_linefile conditionally decrements the line depending on
> +	 whether '\n' is seen.  */

Hmm, should we perhaps suppress the decrement there when linenum is 0? Or
even reject such a use as nonsensical? What buffer_and_nest() inserts is
a positive number in all (correct) cases. "Correct" because line + 1 can
of course wrap, yet I'd again deem such a use bogus. (Really numbers up
from INT_MAX are problematic, as get_linefile_number() takes a pointer to
plain int.) And .linefile isn't intended for use in user code anyway, so
we're pretty free to reject forms we wouldn't generate ourselves.

Jan

> +      if (line_number < 0
> +	  || (line_number == 0 && input_line_pointer[-1] != '\n'))
> +	line_number = physical_input_line;
> +      if (fname != NULL)
>  	abort ();
>        if (next_saved_file == NULL)
>  	fname = physical_input_file;
>