Can someone explain what’s up with the CPU benchmark results on the dpiet pi survey page? I’ll add a link so you don’t have to search
We will have a look
Strange. First I thought that someone might have uploaded invalid data. But we do make sanity checks already, whether it is positive integer numbers etc. I added a further check to assure numbers are not too high either, and all data is good, within 1 and 1000 seconds CPU time.
Also, there were no syntax changes in the script since several versions. The only explanation I am currently investigating is whether bash itself, it’s arithmetic syntax/behaviour has changed, or is bugged. There was a bash upgrade 2024-08-25, but could be also some shared library.
EDIT: Hmm, maybe this recent bug “fix” caused the issue: #1023053 - bash FTBFS on musl: duplicate definition of strtoimax - Debian Bug report logs
Ah nope, here it is: #1078556 - bash: 5.2.21-2.1 to 5.2.21-2.1+b1 breaks printf %.2f .1 - Debian Bug report logs
We indeed use printf '%.2f' to transform the “hectoseconds” back to seconds with 2 decimal places. As of above bug report, this feature of bash’s printf is currently bugged, causing the issue. Hmm, I’ll try to find a temporary workaround.
EDIT: Fixed 
1 Like
This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.