Re: Random panic in load_balance() with 3.16-rc

[Jakub Jelinek <jakub@redhat.com>]

On Sat, Jul 26, 2014 at 10:20:55PM +0200, Markus Trippelsdorf wrote:
> On 2014.07.26 at 15:55 -0400, Theodore Ts'o wrote:
> > On Sat, Jul 26, 2014 at 09:35:57PM +0200, Markus Trippelsdorf wrote:
> > > 
> > > But fortunately the workaround for the new inode.c bug is the same as
> > > for the original bug: -fno-var-tracking-assignments. 
> > > 
> > > It would make sense to enabled it unconditionally for all debug
> > > configurations for now.
> > 
> > What's the downside of enabling this unconditionally on a compiler
> > with the bug fixed?  I assume a certain amount of optimization will
> > lost, but is it significant/measurable?
> 
> Only the quality of the debug info would suffer a bit.

Which for various tools that use kernel's debug info is a significant
difference.
So adding the option even for fixed gcc is undesirable (and, tracking
gcc version numbers only is not enough, I guess most of the distro gccs
will backport the fix soon).

This PR is the first -fcompare-debug wrong-code in the last few years
I remember. There are -fcompare-debug failures from time to time, but
usually they are just that either there is insignificant code change or
no change at all, just changes in the text dump files -fcompare-debug
uses to check whether there might be code differences or not.
GCC's stated goal is that -g should not affect code generation, so we
treat all such differences as bugs, but most of the time they aren't
breaking anything.

	Jakub