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

[Andrew_Pinski at PlayStation dot Sony dot Com]

Mike Stump <mrs@apple.com> wrote on 01/04/2007 02:36:26 PM:

> On Jan 4, 2007, at 12:59 PM, Andrew_Pinski@PlayStation.Sony.Com wrote:
> > One: you cannot turn off the warning without turning off all warnings.
> 
> And is that a feature, or a bug?  Sounds kinda like a feature to me.

It would be a feature if the user could fix their code if it was broken
but their code is not the issue here but GCC is broken.  So users get a
warning which is bogus to them and not helpful at all.  What they will do
is just move to an older version of GCC and give up on the newer ones. 
This
is why this needs to be an internal error and not a warning.  I don't 
think we
really need to warn at all even without checking enabled.
 
> > Two: when you turn off all the warnings, you still get gabarbe to the
> > screen.
> 
> I can live with that.  If someone wanted to conditionalize the two 
> tree dumpers on warnings coming out, that'd be better of course.

Considering this is all internal to GCC rather than an issue with people's 
code.
I rather not warn at all because people might think their code is broken 
when
instead GCC is broken.

The real question is why was this a warning with junk after it in the 
first place,
it seems wrong to be a warning which the user is not able to resolve at 
all.



To an user this warning is not helpful at all since it says nothing about 
their code
except GCC is broken internal so what can the user do, nothing except 
report a bug.  They
cannot turn off just that warning.

I am thinking about applying this patch as obvious and then letting people 
file more bugs
about problems internally to GCC.


-- Pinski