Sven Köhler <sven.koeh...@gmail.com> [2023-11-26 17:09:23 +0100]:
> So Pádraig's patch does allow for parsing lowercase k, but it does not
> change numfmt to use lowercase k in its output in si mode.
>
> As Pádraig has shown, ls uses lowercase k in --si mode. So it uses lowercase
> k for 1000. I think that numfmt should behave the same for consistency
> reasons.
>
> Also, the texinfo then seems incomplete. It lists kB but not k alone, as
> used by ls in --si mode.
>
Danger Will Robinson... The situation w.r.t. B vs. B-less is more complicated
than that; the texinfo doc is (and has been for years) self-contradictory
regarding the semantics of B vs. B-less suffixes. It just depends which
section you happen to refer to.
This thread from a few years back lays out the B-vs.-no-B confusion in
gory detail:
https://lists.gnu.org/archive/html/coreutils/2020-09/msg00001.html
NOTE: The above was posted in 2020 when coreutils 8.32 was current; the
texinfo sections referred to in the initial message of that thread have
changed: In particular, the section referred to in the above thread as
"Section 2.3 "Block size" is now Section 2.2 of the present (coreutils 9.4)
texinfo.
Glenn
