Re: SIGILL in libcrypto.so.14 on NetBSD 9.0 PinebookPro

[Yorick Hardy <yorickhardy%gmail.com@localhost>]

Dear Piotr,

(Apologies for the top-posting). I can reproduce the SIGSEGV
on amd64. Does it work if you change your TERM setting? If I export
TERM=xterm then mutt seems to work here.

On 2020-07-08, Piotr Kolasiński wrote:
> Yes - I've continued in gdb and two times more received SIGILL, but later SIGSEGV.
> That is why I started mutt in gdb, because the final error is SIGSEGV.
> But this happens only when I run mutt in pure st terminal emulator (from
> suckless.org). This not happens when I run it in xterm and (what is more
> interesting) in st with tmux running.u
> So I will dig more and try catch where the problem exists.
> 
> Thanks
> Piotr
> 
> On Wed, Jul 08, 2020 at 08:32:09PM +0200, Martin Husemann wrote:
> > On Wed, Jul 08, 2020 at 08:21:18PM +0200, Piotr Kolasi?ski wrote:
> > > Hi
> > > 
> > > I try to run mutt on PinebookPro NetBSD 9.0 and receive SIGILL.
> > > When I start it in gdb, it points to libcrypto.so.14 function
> > > _arm8_sha512_probe. The name suggest some checks, and $pc register point
> > > to sha512su0, which is probably not supported on this chip. I found that
> > > it exists only if ARMv8.2-SHA is implemented. I'm wondering is it
> > > mistake in compiler settings or maybe in the library dynamic checking of
> > > this feature should be implemented.
> > 
> > This is normal openssl probing for cpu features, it should catch the
> > signal and not use the asm code using that instructions.
> > 
> > Just continue in gdb and all should be fine.
> > 
> > Martin

-- 
Kind regards,

Yorick Hardy