[PATCH][RFC] Another approach to fixing "bad" inlining

[Richard Guenther]

On Tue, 10 Jul 2007, Dave Korn wrote:

> On 10 July 2007 16:01, Richard Guenther wrote:
> 
> > On Tue, 10 Jul 2007, Dave Korn wrote:
> > 
> >> On 10 July 2007 15:23, Richard Guenther wrote:
> >> 
> >> 
> >>   Excuse me for butting-in on what is a side-issue, but ...
> >> 
> >>> where this is undefined at runtime only, so we may not issue an error.
> >> 
> >>> Now we will happily inline, but simply leave y uninitialized (which
> >>> should be valid, as the code is undefined at runtime).
> >> 
> >> ... what is "undefined at runtime only"?  I couldn't find a definition for
> >> this distinction between different kinds of undefined behaviour, and the
> >> note at n2794/3.18.2 suggests there's no reason not to issue an error.
> >> 
> >>   Or is this part of the great linguistics debate about 'may' vs. 'can'?
> >> :-) 
> > 
> > I was basing my knowledge on
> > 
> > http://gcc.gnu.org/ml/gcc-patches/2004-10/msg00772.html
> > 
> > so "undefined at runtime only" just means we may not issue an error.
>  
> 
>   Ah, so I think I would be correct in reading your earlier post as:
> 
> " [ ... ] where this is undefined, but we don't find out until runtime, so we
> may not (in the sense of 'we are not able to', not in the sense of 'we are
> forbidden from') issue an error (because we don't know that we should) [ ...
> ]"
> 
> ... yes?  (Just asking for clarity's sake.)

I believe it's 'we are forbidden from'.

Richard.