Tiny Core Extensions > TCE Bugs

sfwbar has futex-related crashes on TCL14

(1/1)

GNUser:
I'm on TCL14 x86_64, experimenting with wayland. Everything works perfectly except that  sfwbar  randomly becomes completely unresponsive, requiring SIGKILL. Crashes sometimes happens within 5 minutes of booting, sometimes after an hour or more. I have not figured out how to provoke a crash.

I've reported this issue to the  sfwbar  developer, but he seems unable to reproduce the problem. I think I need to exclude something TCL14-specific (e.g., kernel, python, or some library version) being the issue.

When I run  sfwbar  with  strace  and redirect output to a file, the last line in the file (just before my SIGKILL) is always this:

--- Code: ---futex(0x7fecb17f1c40, FUTEX_WAIT_PRIVATE, 2, NULL) = ?

--- End code ---
futex appears elsewhere in the strace the but the = ? part only appears as part of the last line, which makes me think that it is the cause of the crash.

Does seeing futex ... = ? in strace cast suspicion on anything specific that you can think of?

P.S. If you'd like to see some complete straces, here you go:
  a. strace of a crash: https://gnuser.ddns.net/public/log1.txt
  b. strace of another crash: https://gnuser.ddns.net/public/log2.txt
  c. strace of sfwbar running for an hour without crashing: https://gnuser.ddns.net/public/okay.txt

GNUser:
futex is a kernel system call. I scoured the internet and it seems that, at least some of the time, futex-related hangs are associated with python code. sfwbar does use python3 for some of its widgets.

I'm going to disable all the widgets I don't really need. That will only leave a total of 4 widgets, 1 of which relies on python. I'll try using python3.9 instead of 3.6 for the widget that needs it. We'll see how it goes.

GNUser:
Problem was the bluetooth widget, which crashes on a system that does not have bluetooth enabled. I alerted the sfwbar developer. Issue is solved.

Navigation

[0] Message Index