Re: [PATCH, C++] Make Canonical ICE instead of just warn when mismatching

[Andrew Pinski <pinskia at physics dot uc dot edu>]

> 
> On Thu, Jan 04, 2007 at 03:42:36PM -0800, Andrew_Pinski@PlayStation.Sony.Com wrote:
> Do not do that.
> 
> A patch which draws immediate disagreement is _NEVER_ obvious, no
> matter how thoroughly you ignore the response.

Except what Mike is talking about is not really an issue we should
be worrying about right now because again this is an internal check.
If GCC is broken, why try to continue?  This is just like any other
internal check inside GCC we have.  Do you want to continue to compile
when a RTL instruction is invalid for a target and then get wrong code
which is harder to debug?  I don't think so.  Then why have a warning
which is not really helpful at all to users saying GCC is broken and
might miscompile their programs.  Really this should have been an 
internal compiler error to begin with and not a warning.

Also what Mike is talking about is doing a compile farm for a world build
which is a different issue and seperate issue.  It should not be an issue
with a patch which actually just corrects a mistake of generating a bogus
warning.


Again the warning is useless to the user as it is an internal GCC issue
rather than a problem with their code.  The warning should be turned into
an internal error as it is an internal check and not a problem with the
source that is being compiled.  warning should only be used for problems
with the source being compiled and not internal GCC problems, those should
be internal errors.

Also I am about to fix more internal problems and then post more internal
checkers do you want a warning or an error?  Warnings like this are all but
useless to the user.  Maybe we should have -Winternal-checking but that seems
like a stupid idea since the internal problems need to be fixed inside GCC
and not in the sources that are being compiled.

-- Pinski