Re: [Nano-devel] Kernel panic on Ubuntu 1610 with NFS - nano specific

[Benno Schulenberg]

On Sun, Oct 16, 2016, at 15:45, Darren Austin wrote:
> It's reproducable with --ignore, and with .nanorc and /etc/nanorc
> moved out of the way.

Okay.  During further testing keep using --ignore.

> I'm happy to do a couple of git builds to test, but i'm not familiar with 
> the bisect command, so you may need to give specific instructions... :)

Okay.  First read README.GIT, and do the things mentioned
there, up to ./autogen.sh.  You may need to install some of
the things under Prerequisites in order for autogen.sh to
run successfully.  For further instructions, see below.

> Do you still need me to do test builds since the bug still occurs
> even with --ignore?

Yes.  If it hadn't occurred with --ignore, we would have had
a workaround and bisecting wouldn't be as urgent.

But to be clear: you have found a bug in the kernel, not in
nano.  No matter what nano does, it should be *impossible*
for it to bring down the kernel.  The kernel may kill nano
(if nano does something entirely stupid), but the other way
around is a capital fault in the kernel.

> Isn't locking off by default?

It should be, but in a previouis version Ubuntu had it
switched on by default in /etc/nanorc.

> By moving the .nanorc and /etc/nanorc out 
> of the way i've rever[t]ed it to defaults,

Yes, so that's okay.  Keep them out of the way, or
always use --ignore during testing.


Before bisecting, and after ./autogen.sh, first run
  ./configure && make
and then run 'src/nano --ignore ...' and follow your
reproduction recipe.  If that causes a panic, then run
  ./configure --enable-tiny && make
and try your recipe again.

If that does /not/ cause a panic, then stop there.

If it does cause a panic, then run
  git checkout 8edb0968
  ./configure  &&  make
and try your recipe (always using src/nano --ignore).

If it panics, then run
  git bisect reset
  git bisect start
  git bisect bad
  git checkout 162d534f  &&  make
and try your recipe again.

This checkout should not panic, as it equals the 2.5.3 version
of nano that runs on Ubuntu 16.04.  If, however, it /does/
panic, you can stop here.

If it does not panic, run
  git bisect good
  make

And just follow what bisect tells you, reporting
'git bisect good' if it doesn't panic, and
'git bisect bad' if it panics.  After each good or bad,
run make again to compile the new checkout that
bisect has made.  Continue until bisect reports
which is the first bad commit.

Benno

-- 
http://www.fastmail.com - Accessible with your email software
                          or over the web