[Toybox] [PATCH] taskset: Document and add test for 0 PID

[Rob Landley]

On 7/3/25 10:12, Jesse Rosenstock wrote:
> Note that toybox and util-linux show different output.  I'm not sure if
> this should be considered a bug or not.
> 
> % ./toybox taskset 1 ./toybox taskset -p 0
> pid 0's current affinity mask: 1
> % ./taskset 1 ./taskset -p 0
> pid 2140985's current affinity mask: 1

I'm assuming the one you're running out of the current directory is 
util-linux instead of the toybox version. Why you're running the 
util-linux version out of the current directory instead of the $PATH I 
couldn't tell you. But on my debian-ish system:

$ toybox taskset 1 toybox taskset -p 0
pid 0's current affinity mask: 1
$ taskset 1 taskset -p 0
taskset: bad usage
Try 'taskset --help' for more information.
$ taskset --version
taskset from util-linux 2.38.1

I'm happy to change the behavior here, but am unclear on what the right 
behavior is. Something I hadn't noticed when I applied your patch was:

$ TEST_HOST=1 make test_taskset
scripts/test.sh taskset
taskset: bad usage
Try 'taskset --help' for more information.
FAIL: taskset set mask to all
echo -ne '' | '(taskset $MASK taskset -p 0 | sed "s/.*: //")'
--- expected	2025-07-05 14:15:56.329786399 -0500
+++ actual	2025-07-05 14:15:56.329786399 -0500
@@ -1 +0,0 @@
-f
make: *** [.singlemake:929: test_taskset] Error 1

The host no longer passes those tests on my laptop.

(Always fun when there's no specification, just a magic implementation 
we have to use as reference where "whatever it does must be the correct 
behavior, and then THAT behavior is inconsistent.)

Do you know which util-linux version changed util-linux's behavior for 
this command?

Rob

P.S. Filtering host tests so the ones known not to pass on a specific 
host environment don't get run is an issue that's come up a lot before, 
and I'm not sure there IS a good general purpose solution. There's a lot 
of checks for macos, but bash is also full of version skew over the 
years, ls changed its default output-to-tty from -q to -b a few years 
back (so we had to do commit 007af3537d18 to match the new one), Elliott 
got bit last year by od changing the whitespace in its output (commit 
f4ab441ca363)... Testing is HARD.