Re: [PATCH 04/11] Fix expansion point loc for macro-like tokens

[Jason Merrill <jason at redhat dot com>]

On 04/25/2012 05:07 AM, Dodji Seketeli wrote:
> +	  /* If the first token we got was a padding token, let's put
> +	     it back into the stream so that cpp_get_token will get it
> +	     first; and if we are currently expanding a macro, don't
> +	     forget that information.  */
> +	  cpp_hashnode *macro =
> +	    (pfile->context->tokens_kind == TOKENS_KIND_EXTENDED)
> +	    ? pfile->context->c.mc->macro_node
> +	    : pfile->context->c.macro;
> +	  _cpp_push_token_context (pfile, macro, padding, 1);

What about the other places that call _cpp_push_token_context with a 
NULL macro argument?  Don't we want to continue the current macro 
context in that case, too?  Perhaps we should move this new code inside 
_cpp_push_token_context for the case when the macro parameter is NULL.

Jason