Closed XeonBloomfield closed 4 years ago
Interesting but. Thanks.
The problem is that at the moment there is poor support for the number of sensors changing during the operation. Basically we assume that the number of sensors on init
will remain the same, but interestingly, the GPU is being toggled on and off depending on the running application.
This might take some time to fix, to properly be ready for changing sensors per measurement.
Right now version 1.0.0b (as well as the latest code from master branch) is unusable on that hardware when this situation occurs, however 0.8.3 works great but lacks new features.
What comes to my mind as a quick fix for this problem:
This prevents s-tui from crashing and gives time to do handling of changing sensors per measurement later. If you like that solution, I can do pull request with implementation.
0.8.3 should be working, as it only has a single sensor measured for each category. So this makes sense. I think your solution is a good one for a quick fix. For the longer term, we should move away from using a list for the subgraphs and use dictionaries instead. That way only entries with existing values can be populated, with an optional button to refresh the entire view. Please feel free to open a PR, and ask anything I might assists you with.
Thanks @HughWarrington. Yes this is related. s-tui does not handle a change in the number of sensors during operation gracefully. I'll do my best to look into it promptly as it appears more widespread that I initially thought.
@HughWarrington, I have pushed an update that should fix this bug.
There is more work to do, but it should at least prevent the crashes.
Can you please try running from source or installing version 1.0.1
from pip and see if it fixes the issue for you?
@amanusk Thanks for the amazing work squashing thread crashes. Just tested 1.0.1
from pip and it's been very stable in handling sensors hot swap operations.
@amanusk many thanks for the quick response. I've tried running 1.0.1 from pip, and no problems so far after running for a day in the same environment as before.
Step 1: Describe your environment
Step 2: Describe the problem:
Observed Results:
Problem occurs on ThinkPad with Intel and Radeon GPU. When you start s-tui before running any application on dedicated GPU (when it's temperature is not being reported -
sensors
output attached in problem reproduction part) and then start using that card s-tui crashes. Same problem occurs vice versa - if I start s-tui with Radeon GPU active, I can see it's temperature graph and crashes after closing application using dedicated GPU.Debug Results, output of
s-tui -d
created in a file_s-tui.log
:Step 3: Reproduce the problem:
Steps to reproduce:
sensors
)Sensors output with Radeon GPU disabled:
Sensors output with Radeon GPU enabled:
If you need any help resolving this issue, let me know.