Date: Thu, 16 Feb 2023 19:21:13 +0100 From: Mateusz Guzik <mjguzik@gmail.com> To: Mark Millard <marklmi@yahoo.com> Cc: dev-commits-src-main@freebsd.org Subject: Re: git: d9d5f2c042a5 - main - cpuset: add --count [changed/broken command line parsing] Message-ID: <CAGudoHHZ64291yk_y7fERsgaRtS%2BGTGYmye6ynKOh%2BjUbfj97g@mail.gmail.com> In-Reply-To: <EDD615B9-0F47-4D9B-8CD8-A5623AF08293@yahoo.com> References: <D2D940CD-B1A0-41B7-8C02-63CA63940525.ref@yahoo.com> <D2D940CD-B1A0-41B7-8C02-63CA63940525@yahoo.com> <CAGudoHETv8i9pWS=x0sx%2B5TGLnWwA%2BsjmpD%2BOC5OZ9J9rThtVw@mail.gmail.com> <EDD615B9-0F47-4D9B-8CD8-A5623AF08293@yahoo.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 2/15/23, Mark Millard <marklmi@yahoo.com> wrote: > On Feb 14, 2023, at 23:08, Mateusz Guzik <mjguzik@gmail.com> wrote: > >> On 2/15/23, Mark Millard <marklmi@yahoo.com> wrote: >>> Mateusz Guzik <mjg_at_FreeBSD.org> wrote on >>> Date: Sat, 04 Feb 2023 17:51:27 UTC : >>> >>>> The branch main has been updated by mjg: >>>> >>>> URL: >>>> https://cgit.FreeBSD.org/src/commit/?id=d9d5f2c042a51a9f0dd69eb1fc349efd81ffa483 >>>> >>>> commit d9d5f2c042a51a9f0dd69eb1fc349efd81ffa483 >>>> Author: Mateusz Guzik <mjg@FreeBSD.org> >>>> AuthorDate: 2023-02-04 17:47:41 +0000 >>>> Commit: Mateusz Guzik <mjg@FreeBSD.org> >>>> CommitDate: 2023-02-04 17:50:41 +0000 >>>> >>>> cpuset: add --count >>>> >>>> Can be used to count the number of hardware threads in the cpu set. >>>> >>>> For example: >>>> $ cpuset -g -p $$ >>>> pid 2440 mask: 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, >>>> 16, >>>> 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, >>>> 35, 36, 37, 38, 39 >>>> pid 2440 domain policy: first-touch mask: 0, 1 >>>> $ cpuset -g --count -p $$ >>>> 40 >>>> >>>> The intent is to replace calls to sysctl hw.ncpu and kern.smp.cpus >>>> which >>>> can be found in the tree, which are not adequate given existence of >>>> cpusets. >>>> >>>> Right now only -g -p combination is supported to reduce complexity. >>>> As anything else errors out, this can be expanded later as needed. >>>> >>>> Differential Revision: https://reviews.freebsd.org/D36351 >>> >>> This broke long standing commands sequences in some makefiles >>> that I use to build and run some benchmarks, the commands >>> at issue involve use of cpuset. >>> >>> However I can show the oddity in a much simpler context: >>> >>> # cpuset echo "text" >>> text >>> >>> # cpuset echo "-text" >>> usage: cpuset [-l cpu-list] [-n policy:domain-list] [-s setid] cmd ... >>> cpuset [-l cpu-list] [-n policy:domain-list] [-s setid] -p pid >>> cpuset [-c] [-l cpu-list] [-n policy:domain-list] -C -p pid >>> cpuset [-c] [-l cpu-list] [-n policy:domain-list] >>> [-j jailid | -p pid | -t tid | -s setid | -x irq] >>> cpuset -g [-cir] >>> [-d domain | -j jailid | -p pid | -t tid | -s setid | -x >>> irq] >>> cpuset -g --count -p pid >>> >>> >>> By contrast, in my stable/13 context, so, showing the old behavior: >>> >>> # cpuset echo "-text" >>> -text > > FYI: the use of quotes was in my original text for other > reasons. I should not have duplicated them into the above > examples. (Same results without quotes.) > >> the idiomatic way to pass progs to spawn is to denote the end of any >> opts with -- > > The cpuset man page gives no indication that it requires such > notation. Programs with that requirement normally document it > explicitly. (Programs that do not require the ntoation often > allow the notation without documenting it.) > > There may be a consideration of how long it has allowed such > a lack of use of "--" and how many things might break because > it worked before. > > Lots of programs taking commands do not require such. For > example: "time echo -text", "truss echo -text", and "ktrace > echo -text". These have not and are not being changed to > require such. They do allow the "--" notation. > > > If it stays as it is, the change should probably be in UPDATING > or other such place(s) that note where changes are required to > keep things working. > >> so in particular you would: >> cpuset -- echo -text > > Yep. > >> the fact that previous version worked without it is kind of a bug > > I disagree with that last statement, given the many programs > that do not require such. Some programs have some specific > notational complication lead to making it required. But, from > what I've seen, most of the others do not make "--" required. > > So I found that on Linux 'perf stat' et al don't require --. Due to a just found bug in the change anyway I decided to just whack it: https://cgit.FreeBSD.org/src/commit/?id=2924ae099a1e8ffc5b6040f3af5645aa54c0f629 -- Mateusz Guzik <mjguzik gmail.com>
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAGudoHHZ64291yk_y7fERsgaRtS%2BGTGYmye6ynKOh%2BjUbfj97g>