Re: g++.old-deja/g++.other/anon7.C

[Per Bothner <pbothner at apple dot com>]

Mark Mitchell wrote:
> This test seems to be emitting extra, invalid, errors in the wake of
> these patches:
>
> 2003-07-22  Per Bothner  <pbothner@apple.com>
>
> 	* line-map.c (add_line_map):  Handle invalid LEAVE request.
> 	Fixes PR preprocessor/11361.
>
> 2003-07-22  Per Bothner  <pbothner@apple.com>
>
> 	* diagnostic.c.(diagnostic_report_current_module):  Update to match
> 	2003-06-05 changes to push_srcloc and pop_srcloc.
>
> Would you mind taking a look to see if you can see what's going on
> there?

What's happening is that pushdecl_class_level is calling push_srcloc and 
pop_srcloc which add the pushed DECL_SOURCE_FILE and _LINE to the 
#include file stack used for diagnostics.  My patch fixes the 
diagnostics code, so the entire include file stack is shown.

My June 6 change caused part of the include file stack to not get 
printed in error messages.  Before that anon7.C should get spurious 
error messages, because diagnotic.C was misusing push_srcloc/pop_srcloc. 
 So you edited the test case on June 20 to "Remove spurious error 
messages" because we were no longer getting them.  And now they're back.

So you have to decide:  What error messages do you want?  If you don't 
want the "spurious errors" don't use push_srcloc/pop_srcloc.  I suggest 
replacing the calls by:

  location_t save_location = input_location;
  input_location = DECL_SOURCE_LOCATION (f);
  ...
  input_location = save_location;

--
	--Per Bothner
per@bothner.com   pbothner@apple.com  http://per.bothner.com/