Andrew Cagney wrote:
> Jeff Johnston wrote:
> 
>> Andrew,
>>
>>   The following updated patch implements your suggestion of having a 
>> new error routine that throw RETURN_QUIT.  The backtrace command is 
>> now protected with a catch_errors call.  I did not change other 
>> existing calls to catch_errors to use RETURN_MASK_ERROR instead of 
>> RETURN_MASK_ALL.
>>
>>   Ok or is there something I have missed?
> 
> 
> No.
> 
> With regard to s/RETURN_MASK_ALL/RETURN_MASK_ERROR/ that while more 
> correct gives me the willies, so yes we can worry about that later.
> 
> I need to think about s/RETURN_QUIT/RETURN_FATAL/.
> 
> I'm just trying to decide which of:
>     fatal_error OR fatal
> and
>     vfatal_error, fatal_verror, and vfatal
> are better.  Having seen the code, I kind of prefer the shorter "fatal" 
> and "vfatal".  Ok with that.
>

Ok, change made and attached patch has been committed.

2004-11-05  Jeff Johnston  <jjohnstn@redhat.com>

         * defs.h (fatal, vfatal): New function prototypes.
         * stack.c (backtrace_command_stub): Stub to call backtrace_command_1
         via catch_errors.
         (backtrace_command): Change to call backtrace_command_stub via
         catch_errors instead of calling backtrace_command_1 directly.
         (backtrace_full_command): Ditto.
         * utils.c (error_stream_1): New static function.
         (verror): Change to call error_stream_1 instead of error_stream.
         (error_stream): Call error_stream_1 with RETURN_ERROR argument.
         (vfatal, fatal): New functions.


> This should be mentioned in NEWS, I guess with "thread apply all bt" now 
> "works".
>

I'll post a change for this as well as a test case I have.

-- Jeff J.