[racket] gc crash on raspbian

From: Matthew Flatt (mflatt at cs.utah.edu)
Date: Wed Jul 2 04:28:19 EDT 2014

Is it enough to run the web server in its default configuration, or do
you have to run something specific?

Meanwhile, I'll looking into Deren's problem, which seems to be
JIT-related. Just to make sure, does setting the environment variable

 PLTNOMZJIT=y

before running the web server (to disable the JIT) change the result?

Thanks!

At Wed, 02 Jul 2014 09:00:58 +0100, David Griffiths wrote:
> Hi all,
> 
> With Franco's help I have 6.0.1 on the Pi, still crashes, but now looks
> like this:
> 
> Seg fault (internal error during gc) at 0xb46ac000
> SIGSEGV SEGV_ACCERR SI_CODE 2 fault on 0xb46ac000
> Aborted
> 
> cheers,
> 
> dave
> 
> 
> On 01/07/14 13:10, Franco Raimondi wrote:
> > I managed to compile v6.0 on Raspbian, let me know if you are interested and 
> I can share it with you via dropbox or something similar.
> > 
> > Franco
> > 
> > 
> > On 1 Jul 2014, at 13:05, David Griffiths <dave at pawfal.org> wrote:
> > 
> >> Hi Matthew,
> >>
> >> Just to report back, I just upgraded to v5.3.5 on the Pi (after a 48hr
> >> build :) and still getting this problem.
> >>
> >> I'll try and narrow it down - the SQLite api is the first thing to
> >> isolate I guess...
> >>
> >> cheers,
> >>
> >> dave
> >>
> >> On 02/05/14 22:58, Matthew Flatt wrote:
> >>> Version 5.2.1 seems like a long time ago (only two years? really?), and
> >>> some GC bugs have been fixed since then. Is upgrading to v6.0 possible?
> >>> We're offering a free ARM JIT with every upgrade!
> >>>
> >>> While 7 would be surprisingly close to a valid value for a random
> >>> `int`, it turns out that the object header has only 3 bits to hold that
> >>> value, so 7 is actually as far away as possible from a valid value.
> >>>
> >>> At Fri, 02 May 2014 22:44:43 +0100, David Griffiths wrote:
> >>>> Hi Matthew,
> >>>>
> >>>> Sorry, I was a bit short on details, I'm running 5.2.1, also using
> >>>> sqlite via jaymccarthy/sqlite:5:1/sqlite on planet.
> >>>>
> >>>> I had assumed it would be some kind of memory corruption, but "7" looks
> >>>> a bit more specific than I'd expect for that (like some kind of specific
> >>>> error value as it's one more than the PAGE_X types)...
> >>>>
> >>>> I'll try and narrow it down a little, e.g. without sqlite and see what
> >>>> happens.
> >>>>
> >>>> cheers,
> >>>>
> >>>> dave
> >>>>
> >>>> On 02/05/14 19:39, Matthew Flatt wrote:
> >>>>> That means something has gone generally wrong with memory management.
> >>>>>
> >>>>> Do you have a way to replicate the crash? I tried simply running the
> >>>>> web server, but I imagine there's a little more to it.
> >>>>>
> >>>>> At Fri, 02 May 2014 14:09:01 +0100, David Griffiths wrote:
> >>>>>> Hi all,
> >>>>>>
> >>>>>> I'm running a racket web server on a Raspberry Pi, and occasionally
> >>>>>> getting aborts with:
> >>>>>>
> >>>>>> Unhandled info->type 7
> >>>>>>
> >>>>>> Which I've tracked down to src/racketgc2/newgc.c
> >>>>>> Any pointers how what this could be and how to fix/work around it?
> >>>>>>
> >>>>>> cheers,
> >>>>>>
> >>>>>> dave
> >>>>>> ____________________
> >>>>>>  Racket Users list:
> >>>>>>  http://lists.racket-lang.org/users
> >>>>>
> >>> ____________________
> >>>  Racket Users list:
> >>>  http://lists.racket-lang.org/users
> >>>
> >>
> >> ____________________
> >>  Racket Users list:
> >>  http://lists.racket-lang.org/users
> > 
> > 
> > ____________________
> >   Racket Users list:
> >   http://lists.racket-lang.org/users
> > 

Posted on the users mailing list.