Re: [PATCH] PR c/10175: Fix for -Wunreachable-code

[Geoff Keating <geoffk at geoffk dot org>]

Mark Mitchell <mark at codesourcery dot com> writes:

> On Mon, 2003-03-31 at 11:42, Steven Bosscher wrote:
> > Op ma 31-03-2003, om 21:19 schreef Mark Mitchell:
> > > 
> > > In my opinion, to do this right, the code needs to be moved further into
> > > the front end.  On the other hand, to do it the way we have it now,
> > > whatever is causing jump not to issue the warnings it used to should be
> > > figured out and fixed.
> > 
> > Moving the warning to the front end is not really an option if you wish
> > to preserve the current behavior.  never_reached_warning is (RTL) CFG
> > based, and we don't know anything about the CFG in the front end.
> 
> Yes, I know.  And I think that's broken.  (See, for example, my point
> about code for which RTL is never generated.)

We could certainly also emit warnings in the front ends, but the back
end also detects unreachable code and should issue the warning when it
does so.  (The cases that the backend detects are often the most
interesting from the programmer's point of view.)

-- 
- Geoffrey Keating <geoffk at geoffk dot org>