[Vm-dev] Re: Cog Spur Linux pthread error?

Eliot Miranda eliot.miranda at gmail.com
Mon Jun 23 01:35:20 UTC 2014


On Sun, Jun 22, 2014 at 2:24 PM, Frank Shearar <frank.shearar at gmail.com>
wrote:

> On 22 June 2014 19:40, Frank Shearar <frank.shearar at gmail.com> wrote:
> > On 21 June 2014 19:06, Eliot Miranda <eliot.miranda at gmail.com> wrote:
> >> Hi Frank,
> >>
> >>     (Bob, thanks for your reply).
> >>
> >>
> >> On Fri, Jun 20, 2014 at 2:39 PM, Frank Shearar <frank.shearar at gmail.com
> >
> >> wrote:
> >>>
> >>> Hi Eliot,
> >>>
> >>> I thought I'd take a stab at setting up a CI job for Spur VM+Image,
> >>> and after futzing around with filenames, got an actual execution of
> >>> the VM. And then things broke. The full output's here:
> >>>
> >>>     http://build.squeak.org/job/SqueakTrunkOnSpur/6/console
> >>>
> >>> but the error itself is this:
> >>>
> >>>     pthread_setschedparam failed; consider using ITIMER_HEARTBEAT:
> >>> Operation not permitted
> >>>
> >>> and the process exited with status 22.
> >>>
> >>> Is this a simple thing to fix? Any idea what hoops I need to jump
> >>> through to work around the problem?
> >>
> >>
> >> So this is always described in the README on my website, see e.g.
> >> http://www.mirandabanda.org/files/Cog/VM/VM.r3006/README.3006.  I
> *could*
> >> put checks in the VM that saw if the kernel was >- 2.6.13 and whether
> >> /etc/security/limits.d/VM.conf existed and issue a more helpful error
> >> message.  But this is work and I feel like the README should be
> sufficient.
> >> What do y'all think?
> >
> > I should have read the readme.
> >
> > I do have an additional newbie question though: the readme says "log
> > out and back in again", which is fine for a user. In my case I'm not a
> > user, I'm a Jenkins service. I thought the equivalent would be
> > "restart Jenkins", which I did, and has apparently been insufficient.
> > Anyhow, _as a user_ I can now run cogspur on build.squeak.org. I'll
> > fight with Jenkins a bit more and see if I can get _Jenkins_ to do so.
>
> Ah, silly me. Of course, what you need is a process restart, which
> means stopping the service, and starting it again. So now we have a
> running Cog Spur! It's also showing something interesting, in the form
> of a segfault at
> http://build.squeak.org/job/SqueakTrunkOnSpur/12/console:
>
> Smalltalk stack dump:
> 0xbf8e1ad4 M [] in AllocationTest>testOutOfMemorySignal 0x77615df8:
> a(n) AllocationTest
> 0xbf8e1b10 I Interval>collect: 0x776c2a28: a(n) Interval
> 0xbf8e1b30 M [] in AllocationTest>testOutOfMemorySignal 0x77615df8:
> a(n) AllocationTest
>

I'm currently working on a rewrite of the mmap memory management for linux.
 There are a couple of issues, the initial mmap is at too high an address
meaning that there is very little address space to mmap above it (the first
segment contains nil and must be the lowest oldSpace segment), and any
subsequent mmap mustn't conflict with any others, but this isn't correctly
implemented yet, so on linux the current Spur VMs a) can't allocate very
much memory and b) may wrongly reallocate memory used for other facilities.
 Hopefully this should be resolved soon.

(I do my daily development on Mac OS X so finding issues on linux and win32
alas always lags).



> frank
>
> > frank
> >
> >> --
> >> best,
> >> Eliot
>



-- 
best,
Eliot
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.squeakfoundation.org/pipermail/vm-dev/attachments/20140622/ebacf7aa/attachment-0001.htm


More information about the Vm-dev mailing list