[Bug rtl-optimization/67778] [6 Regression] ICE on valid code at -O3 on x86_64-linux-gnu in maybe_record_trace_start, at dwarf2cfi.c:2297

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67778

--- Comment #11 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
(In reply to Jakub Jelinek from comment #10)
> (In reply to Segher Boessenkool from comment #9)
> > I cannot reproduce the problem, #c2, not with any flags I tried,
> > not with a 20151122 compiler either.  Help?
> 
> #c2 testcase, x86_64-linux compiler, -m32 -O2, reproduces this for me
> reliably.
> Say with -mtune=generic -march=i386 (or -march=x86_64, or -march=i686).
> -m32 -O3 too.  r231142 reproduces it too.
> -m32 -O2 -fno-shrink-wrap does not ICE.  Are you using a cross-compiler from
> some other target?

But it reproduces on #c2 also with -m64 -O2 and -m64 -O3.
In all cases it is
Block 10 needs the prologue.
After wrapping required blocks, PRO is now 10
Avoiding non-duplicatable blocks, PRO is now 10
Bumping back to anticipatable blocks, PRO is now 9
Performing shrink-wrapping.
on main, where bb10 is not in any loop, but bb9 is part of a loop.