CoCalc Docker image sagemath kernel crashes

31 views
Skip to first unread message

Matthew Saltzman

unread,
Nov 20, 2024, 3:01:44 PM11/20/24
to CoCalc
I have two CoCalc Docker images installed on two different computers. On one, the sagemath kernel runs fine. On the other, the kernel hangs when attempting to do anything.

While waiting, the status button next to the line number says "Sending to be evaluated using ." and eventually times out. Running sagemath in a terminal results in a long traceback, ending with the lines below.

Any suggestions how to continue troubleshooting?

TIA!

/lib/x86_64-linux-gnu/libc.so.6(+0x29d90)[0x7f41037cfd90]
/lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0x80)[0x7f41037cfe40]
python3(_start+0x25)[0x55c3a460d095]
------------------------------------------------------------------------
Attaching gdb to process id 6637.
Cannot find gdb installed
GDB is not installed.
Install gdb for enhanced tracebacks.
------------------------------------------------------------------------
Unhandled SIGILL: An illegal instruction occurred.
This probably occurred because a *compiled* module has a bug
in it and is not properly wrapped with sig_on(), sig_off().
Python will now terminate.
------------------------------------------------------------------------
/usr/local/sage/src/bin/sage-python: line 2:  6637 Illegal instruction     (core dumped) sage -python "$@"

Charles Taylor

unread,
Nov 20, 2024, 5:41:39 PM11/20/24
to Matthew Saltzman, CoCalc
Illegal instruction probably means that it attempted to use an instruction (e.g. AVX or similar) that your processor isn't capable of, hence the different behaviour on the two different systems. I'm willing to bet the system with the error is older, or otherwise has a less capable CPU.

If you look back in the forums you'll see I ran into something similar a while back, on a very old machine (Mac Pro 1,1, from '06), after trying for a while to compile sage without the problematic instructions, I gave up, so unfortunately I can't be of much help there. Eventually that system was replaced through the natural course of things and it works normally on the newer machine that replaced it.

c



--
CoCalc: Collaborative Calculation in the Cloud -- https://cocalc.com
---
You received this message because you are subscribed to the Google Groups "CoCalc" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cocalc+un...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/cocalc/586c4fc7-e655-4afe-bdd5-b5cf15d291f3n%40googlegroups.com.

Matthew Saltzman

unread,
Nov 21, 2024, 10:30:19 AM11/21/24
to Charles Taylor, CoCalc
That's kind of what I was afraid of.

Yes, the machine that fails is quite old now. We're supposed to be getting updates sometime...

I had built old versions of Sage on it, but it was looking too complex to build and deploy updates after the switch to Jupyter for a UI.

Thanks.
-- 
Matthew Saltzman
School of Mathematical and Statistical Sciences
Clemson University
mjs AT clemson DOT edu
Reply all
Reply to author
Forward
0 new messages