eliot.miranda at gmail.com
Tue Jan 22 18:13:50 UTC 2013
On Mon, Jan 21, 2013 at 10:03 AM, Chris Muller <asqueaker at gmail.com> wrote:
> Hi Eliot, I just wanted to confirm whether these changes are intended
> to address crashing issues of the new 2673 VM.
As David says, no, these are reiterations of changes I made for the 2638 VM.
> I was happy to read
> your release notes of 2673 because they seem to indicate you're aware
> of an issue I experience related to use of bulk becomeForward: -- as
> if not all are getting becomed. However I went back to 2672 for now
> because 2673 crashes very quickly during OSProcess activity.
That's very good to know. That means that my recent changes have caused a
regression. I finally implemented setting the stack page size and per-page
stack limit in a sane way, but your crashes show I still have the
dimensions wrong. If I may I'd like to use you as a guinea pig since you
can reproduce the linux crashes very quickly. But it'll take a little
while to construct the code I need to measure precisely how much stack
space is used. For now using 2672 should be OK provided your bulk becomes
don't create huge numbers of new roots. This is something that tends to
happen a lot in Newspeak, where classes are often new objects, but not so
much in Squeak.
> On Sun, Jan 20, 2013 at 1:23 PM,
> <squeak-dev-noreply at lists.squeakfoundation.org> wrote:
> > Eliot Miranda uploaded a new version of
> VMConstruction-Plugins-OSProcessPlugin to project OSProcessPlugin:
> > ==================== Summary ====================
> > Name: VMConstruction-Plugins-OSProcessPlugin.oscog-eem.41
> > Author: eem
> > Time: 14 December 2012, 4:42:40 am
> > UUID: 0406d3fc-7b8b-4b72-b20a-f254eeb1b893
> > Ancestors: VMConstruction-Plugins-OSProcessPlugin.oscog-eem.40
> > Check for needing sigaltstack properly in setSignalNumber:handler:
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Vm-dev