With Pharo 1.4 (but not with Pharo 1.3) we have a problem leading to the crash of DrGeo (based on Pharo 1.4) on low memory machine. It is clearly a problem related to Pharo 1.4 Any idea why ? Hilaire -------- Original Message -------- Subject: Re: [Sugar-devel] Dr.Geo on XO 1 Date: Sun, 20 May 2012 13:46:01 +0200 From: Bert Freudenberg <bert@freudenbergs.de> To: Sascha Silbe <silbe@activitycentral.com> CC: Ariel Calzada <ariel.calzada@gmail.com>, Hilaire Fernandes <hilaire.fernandes@edu.ge.ch>, Sugar Devel <sugar-devel@lists.sugarlabs.org>, Rafael Ortiz <rafael@activitycentral.com> On 20.05.2012, at 11:00, Sascha Silbe wrote:
Bert Freudenberg <bert@freudenbergs.de> writes:
Nope, that was not it. I compiled with -mtune=geode, makes no difference.
Make sure to set -march=<architecture> to something appropriate. -march tells gcc what instructions it can use (so anything that doesn't support these instructions will fail to run the resulting executable), whereas -mtune only tells it to optimise for a certain processor type (but the executable will continue to run on other processors as long as they support the instruction set specified by -march).
I actually had used "-march=pentiumpro -mtune=geode" as recommended here: http://wiki.laptop.org/go/Geode_LX and even with an additional -O0 it still crashes.
In gdb I see it's segfaulting inside the GC logic. It's really
puzzling this only happens on the XO-1 ...
You could try booting an XO-1.5 with mem=256M (append that to boot-file in /bootpart/boot/olpc.fth) to simulate the memory pressure of an XO-1.
YES! That reproduces the crash on XO-1.5. Good thinking! :) I could also reproduce it in a virtual Ubuntu 12 with 768 MB RAM (!) but no swap. Top reports: Mem: 766204k total, 601588k used, 164616k free, 45624k buffers Swap: 0k total, 0k used, 0k free, 277024k cached but DrGeo still crashes. OTOH, Etoys runs fine using the same Squeak VM on the same system (and on XO-1). So I wonder what the new DrGeo does differently ... - Bert -