PR 23551: why should we coalesce inlined variables?

[Richard Guenther]

On 7/9/07, Daniel Jacobowitz <drow@false.org> wrote:
> On Sun, Jul 08, 2007 at 06:57:35PM -0700, Andrew Pinski wrote:
> > Now the question comes down to this, how many users have complained
> > about memory usage inside GCC compared to them complaining about
> > optimized debugging info?  I saw the first is more complained about,
>
> I was trying to stay out of this discussion but...
>
> That's a lame argument.  People don't complain about GCC's optimized
> debug info as much, because we've created the expectation that it (and
> GDB's support for it) is crap.  Every time someone starts working on
> tools which require better debug info (off hand, I've seen Fred Fish,
> Alexandre, RTH, and myself do this) they have problems GCC's debug
> info.  If the simple ones don't get fixed there's no point filing the
> more complicated ones.
>
> > Now tools that use debugging info to look at values at specific spots
> > should know that they can't always get the most accurate view with
> > optimized code (they never will be able to).
>
> No, pretty much everything a compiler can conceivably do to your code
> is expressible in debug info.  And I gather that some other compilers
> do much better at this than we do.

So improve the way we are collecting information to produce debug info,
but not disable optimizations.  Really, doubling memory usage from
1gb to 2gb for no gain, even if it is on machine-generated code is not
worth it.

Richard.