Re: [PATCH] Speeding up delta by -fatal-errors

[Gabriel Dos Reis <gdr at integrable-solutions dot net>]

Richard Guenther <rguenth@tat.physik.uni-tuebingen.de> writes:

| Gabriel Dos Reis wrote:
| > Richard Guenther <rguenth@tat.physik.uni-tuebingen.de> writes:
| > | Richard Guenther wrote:
| > | > Hi!
| > | > To speed up automatically reducing testcases with delta we don't
| > | > currently have something like -fatal-errors which exit on the first
| > | > error occoured, but we rather keep going.  So would something along
| > | > the simple approach below be acceptable?  (Plus adding the -fatal-errors
| > | > switch and documentation, of course)
| > | | which would be
| > | | 2004-04-06  Richard Guenther  <richard.guenther@uni-tuebingen.de>
| > | |          * commom.opt (fatal-errors): Add it.
| > |          * errors.c (flag_fatal_errors): Define it.
| > |          (error): Check for flag_fatal_errors.
| > |          * flags.h (flag_fatal_errors): Declare it.
| > |          * opts.c (common_handle_option): Add OPT_fatal_errors.
| > |          * doc/invoke.texi (Warning Options): Document -fatal-errors.
| >   1) I doubt it is that simple -- there are parts of the compiler
| > that
| >      don't care about errorcount.  Those are your targets.
| 
| It's just about exiting early if compilation cannot finish
| successfully. And just for the purpose of speeding up regression
| minimization.

But then, you're modifying the wrong error().  See diagnostic.c.

-- Gaby