[Toybox] [PATCH v3] taskset: Add -c cpulist support

[Rob Landley]

On 3/19/26 07:03, Jesse Rosenstock wrote:
> It turns out I'm not subscribed to the list and
> http://lists.landley.net/pipermail/toybox-landley.net/2026-March/030932.html
> wasn't cced to me.  I think I've tried several times to subscribe, but either
> the form is broken, I was never approved, or something else went wrong.

If you weren't subscribed to the list your messages wouldn't be showing 
up in 
http://lists.landley.net/pipermail/toybox-landley.net/2026-March/030933.html 
and I wouldn't be getting copies of them filed to my "toybox" folder 
(which is checking the list-id header added by the list software). Mail 
from non-subscribers is held for manual review...

Which I see an email from jrosenstock at google.com just was. Which is not 
the same as the jmr@ this was sent from? (Now I'm confused.)

In theory mails should be coming to you (unless the "don't send me 
email" checkbox in your membership settings is checked... Ok, I just 
pulled up the admin interface and you're not in there, in _either_ 
address. So how did this email get through? (I didn't manually approve 
it...) Did you subscribe and unsubscribe recently...? (I just did a 
"view source" on the email I'm replying to and yes it has a list-id. 
When someone is unsubscribed from the list due to excessive bounces or 
something I get an email about it...)

I can submit a dreamhost support ticket about this if it's sufficiently 
weird.

> Thanks for the review.  I've sent a new v4 fixing the macro and docs.

I applied your previous one locally and was trying to clean it up here, 
I just get discouraged easily every time I bump into a new 
https://lwn.net/Articles/1062112/ or similar. (Which mostly just 
accumulate because reading them gives me anxiety. I never had anxiety 
before my third bout with covid! It really sucks. It's a pity I don't 
have good enough health insurance to get properly medicated for it. Oh 
well, at least the free-floating kind eventually went away and now it 
needs something to TRIGGER it. Alas, kind of a target rich environment 
since Biden announced he was running for a second term.)

On the bright side, 
https://arstechnica.com/gadgets/2026/03/google-details-new-24-hour-process-to-sideload-unverified-android-apps/ 
wandered by on mastodon today which is at least SOMETHING. (If that had 
been the opening position I'd have come to terms with it by now. Taking 
a year to get there reads a lot like management trying hard to fully 
close up the system and then doing the smallest possible TACO when 
everybody rebelled. Still: it's good news of a sort, don't punish the 
behavior you want to see, take the "win", breathe deeply...)

>> Is "stride" actually needed?
> 
> It's needed for parity with util-linux and also useful for
> hyperthreading, where threads 2n and 2n+1 can be both physical core n.

Hyperthreading was the only case I could think of either, which is weird 
becomes man taskset on debian explicitly says :3 is supported:

  --cpu-list 0-10:2
     is processors #0, #2, #4, #6, #8 and #10. The suffix ":N" specifies
     stride in the range, for example 0-10:3 is interpreted as 0,3,6,9
     list.

Which is not hyperthreading, that's 
https://en.wikipedia.org/wiki/Barrel_processor which I've never seen 
combined with physical SMP. I suppose it _could_ happen, just... 
infrastructure in search of a user again.

> 0-127:2 is going to be more readable than a list of 64 numbers.

Indeed, but is this really userspace's problem? Wouldn't it make more 
sense for the kernel's scheduler to distribute tasks to separate 
physical cores? To be honest, wouldn't it have made sense for a machine 
with 16 physical cores and 16 virtual cores to have the physical ones be 
1-16 and the virtual ones be 17-32 so the _existing_ taskset could have 
easily...

Sigh, arguing with an installed base of stupid, we're stuck with an 
existing model that's got gnu design decisions all over it. I pushed 
what I'd already applied, I still want to clean it up more but just 
don't have the spoons right now.

Rob

P.S. And "stride" is poorly defined! Does 10101001001001 become 
1-5:2,8-14:3 or 1-3:2,5-14:3? It's NOT UNIQUE. A BAD case of gnu, all 
over this one. Ick!