Peter, All four points fixed and pushed to GitHub:
- CPU panel on SPARC: was probing kstat instances 0..N-1 by number, assuming contiguous zero-based ids. Now walks the live kstat chain instead, so it doesn't matter what range the ids fall in. - fs/sys/buf.h collision: alias renamed to btop_fs. - find/getconf NPROCESSORS_ONLN: source discovery now uses Make's built-in $(wildcard ...) instead of find; thread count tries psrinfo first, getconf as fallback. - -static-libstdc++: no longer forced, only added for the Solaris 10 cross-build (S10=true). Tested on an OpenIndiana VM and Solaris 10 x86 VM so far; still need to re-run this batch on bare metal, an old Acer laptop with OI, which I'll get to shortly. Please let me know how it goes on your SPARC machines. Thanks again for the testing. Alex On Tue, Sep 22, 2026, 2:07 p.m. Peter Tribble <[email protected]> wrote: > Alex, > > On Tue, Sep 22, 2026 at 6:51 PM Alex Gorouvein <[email protected]> > wrote: > > > > Thank you for clarification, Peter. I will ensure my next patch > > incorporates this and the three previously mentioned issues. > > Thanks for all the work you're putting in here. > > > Just curious, which SPARC machine do you have? Many years ago I had a > > headless SS5. > > The machine I tested was a T8-1 (64x 5GHz SPARC-M8 CPUs) running Solaris > 11.4, > it's in the gcc compile farm so is always running which makes it > useful for quick tests. > I have a T4-1 and some T5140s at home, which run for a few hours a week. > > > Alex > > > > On Tue, Sep 22, 2026 at 5:12 AM Peter Tribble <[email protected]> > > wrote: > > > > > Alex, > > > > > > Thanks for the fixes! > > > > > > On Tue, Sep 22, 2026 at 2:15 AM Alex Gorouvein <[email protected]> > > > wrote: > > > > > > > > On the blank CPU panel on SPARC — I don't have SPARC hardware to test > > > > against, so if you get a chance, `kstat -p cpu_stat:0` on that > machine > > > > would help me see whether the counter names or layout differ from > what > > > the > > > > collector expects. > > > > > > Ah, I see the problem. There isn't a cpu 0. That's true in general - > the > > > cpu id > > > doesn't have to start at 0, and the list doesn't have to be > > > contiguous. Especially > > > on SPARC, where the cpu id comes from system topology rather than being > > > simply enumerated. On this system, the id is the range 64-127. So the > logic > > > that loops over 0-63 misses everything. > > > > > > -- > > > -Peter Tribble > > > https://www.petertribble.co.uk/ - https://ptribble.blogspot.com/ > > > > > > _______________________________________________ > > > openindiana-discuss mailing list > > > [email protected] > > > https://openindiana.org/mailman/listinfo/openindiana-discuss > > > > > _______________________________________________ > > openindiana-discuss mailing list > > [email protected] > > https://openindiana.org/mailman/listinfo/openindiana-discuss > > > > -- > -Peter Tribble > https://www.petertribble.co.uk/ - https://ptribble.blogspot.com/ > > _______________________________________________ > openindiana-discuss mailing list > [email protected] > https://openindiana.org/mailman/listinfo/openindiana-discuss > _______________________________________________ openindiana-discuss mailing list [email protected] https://openindiana.org/mailman/listinfo/openindiana-discuss
