sh (pid 1081) segfaults for page address 00000000 at pc 00000000

[Cristian Ionescu-Idbohrn]

On Fri, 6 Feb 2009, Cristian Ionescu-Idbohrn wrote:

> On Thu, 5 Feb 2009, Denys Vlasenko wrote:
>
> > On Wednesday 04 February 2009 14:03, Cristian Ionescu-Idbohrn wrote:
> > >
> > > I see what you mean.
> > > 1.12.4 with the named patch applied segfaults :(
> >
> > These are good news! Figuring out where ash broke would be worse.
>
> Thanks for the tips.
> I'm doing that now.
>
> Unfortunately, 1.11.3 started to segfault too :(
> Seems I was not pacient enough while testing.
> That means even more work.  I'll have to go down to 1.10.x, I presume.

Didn't go down to 1.10.x, just yet.  But I found a possible error source.
crisv32/R64 may be at fault :(

The workaround I found this weekend, is to compile with:

	-O2 -fno-frame-omit-frame-pointer

No segfaults using that.  -O0 and -O1 do not cause segfaults either.

I still don't totally rule out a bug in busybox.  Still.  As, to my
knoledge, noone else reported such a bug yet, on any other architecture,
crisv32/R64 seems to be the trigger (unless proven otherwise) for the
segfaults.  I wash my hands clean now ;-)

Compiler experts at Axis will have to dig into this.


Cheers,

-- 
Cristian