stalld version v1.24.1 released

[email protected],[email protected],[email protected],[email protected]
Subject: stalld version v1.24.1 released

Hello RT users,

I just pushed the v1.24.1 update to stalld. This release has a few bug fixes and some minor tweaks for logging and building. You can find this update in the 'main' branch at:

https://git.kernel.org/pub/scm/utils/stalld/stalld.git/
https://github.com/clrkwllms/stalld
https://gitlab.com/rt-linux-tools/stalld

Note that I consider the kernel.org tree as the gold standard. We keep copies at github and gitlab mainly so that people familiar with the PR/MR workflow style can submit changes.

Derek Barbosa, with advice from Crystal Wood, tweaked the Makefile so it wouldn't be so Red Hat-centric. We basically put all that in the package build step where the specfile overrides CFLAGS and LDFLAGS. There's now a help target in the Makefile and some compiler version checks that hopefully will help other builders.

A segfault was found (and fixed) while running with the adaptive or aggressive mains (rather than the default single-threaded main).

A new command line switch (-N/--no-idle-detect) was introduced to skip the idle detection step in a processing iteration, mainly for testing purposes but some of you might find it useful in production.

Some defensive checks were added to print_boosted_info. There is a bit of debate on whether it's better to continue running or just crash right there. Stay tuned...

I commented out a log message reported that idle detection had detected no activity and so nothing was starving. Got tired of seeing it in the log file.

A double-free was avoided by setting the cpu_info->starving vector to NULL when it's been freed.

We rewrote the sched_debug backend parser that pulls info from the debugfs file. The format changed in 6.12 and we missed that. We unified the parsers and hopefully we'll be a bit more change-proof there. Of course none of you saw this because you're all using the modern super-cool queue_track BPF backend, right?

Finally I'm sure that you'll all noticed the notice in some of the commit messages that I used the Claude AI to analyze bits of this code. I've been working on a test-suite and using Claude to generate most of it. While running, the test suite found the segfault and the double free issues. So since I was already there with Claude, I figured it should get some credit.

The test-suite isn't quite ready for prime time though, so I'll push a branch named 'test-devel' up to the git repos if you want to play with it, or even write some more tests!

In any case, I hope this release will be useful to you. Have a nice day!

Clark

The United States Coast Guard
Ruining Natural Selection since 1790