Faster IT

As the Debian maintainer, I'd be happy if you could narrow the issue down to what is causing this. A valgrind callgrind run is probably useful to see where it...

that's all good, @joshuaboniface . `callgrind_annotate` will get you the info where the CPU burns cycles through processing the log file

Just run `callgrind_annotate valgrind-htop.log`. No `valgrind` in this tool. (the `valgrind-htop.log` is created by the `valgrind --tool=callgrind --log-file=valgrind-htop.log htop` command you listed in https://github.com/htop-dev/htop/issues/1211#issuecomment-1476907047 )

ok, calling `callgrind_annotate` on your `callgrind.out.1005067.txt` file yields: ```-------------------------------------------------------------------------------- Profile data file 'callgrind.out.1005067.txt' (creator: callgrind-3.19.0) -------------------------------------------------------------------------------- I1 cache: D1 cache: LL cache: Timerange: Basic block 0 - 205004804 Trigger: Program...

Which means it scales roughly with the amount of ProcessList items shown :-)

We still need an ASCII fallback cf. `htop --no-unicode`

> hint: https://stackoverflow.com/questions/9948987/detect-current-cpu-clock-speed-programmatically-on-os-x The tl;dr version of this is: Not really possible to get anything but the default frequency on MacOS without very low level measurement loops (which would break...

That is intentional to get the block under the `if base_prefix` stanza. Alternatively it could stay one indent less and the `if base_prefix` could be repeated.

hm, let me check that again. We changed a few more things in the course of getting the `base_prefix = "/"` warning fixed. /DLange

We reverted back to `3.0.6` in prod for reasons unrelated to this one, so sorry this update took so long. Hopefully the newly pushed change is better. /DLange