#1035586 coreutils: pr: -n overly narrows the text in the columns? and eats the separator?

Package:
coreutils
Source:
coreutils
Description:
GNU core utilities
Submitter:
наб
Date:
2023-05-05 20:27:05 UTC
Severity:
normal
#1035586#5
Date:
2023-05-05 20:25:17 UTC
From:
To:
Dear Maintainer,

Given
  % pr -2 -w22 -n american.med+.mwl | cat -A
what do you expect the first bit to be?

We know that (per POSIX) -n is equivalent to -n<tab>5;
we know therefore, that each column looks like
  NNNNN<tab>data
since we know we want two columns,
and the minimal separator between them is one blank,
those will be
  NNNNN<tab>data NNNNN<tab>data
  5    0    x   15    0    x
(remember that <tab> is !print),
so given -w22 we can say that
  2*(5+0+x)+1=22
thus
  x=5.5
or, indeed, 5.

So, were we to emulate this, we would expect
  % out/cmd/pr -2 -w22 -n american.med+.mwl  | cat -A
  $
  $
  2023-05-05 22:02 american.med+.mwl Page 1$
  $
  $
      1^Ia/S      57^IAbern$
      2^IA's/I    58^Iaberr$
      3^IAAC      59^Iaberr$
      4^IAache    60^Iabet/$
      5^IAaliy    61^Iabett$
      6^Iaardv    62^Iabett$
      7^IAaron    63^Iabett$
      8^Iabaci    64^Iabett$
      9^Iaback    65^Iabeya$
     10^Iabacu    66^Iabeya$
     11^Iabaft    67^Iabhor$
     12^Iabalo    68^Iabhor$
     13^Iaband    69^Iabhor$
     14^Iaband    70^Iabhor$
     15^Iaband    71^Iabhor$
     16^Iabase    72^Iabhor$
     17^Iabase    73^Iabide$
so what in the world happened to
  % pr -2 -w22 -n american.med+.mwl | cat -A
  $
  $
  2023-05-05 18:58 american.med+.mwl Page 1$
  $
  $
      1^Ia/    57   Ab$
      2^IA'    58   ab$
      3^IAA    59   ab$
      4^IAa    60   ab$
      5^IAa    61   ab$
      6^Iaa    62   ab$
      7^IAa    63   ab$
      8^Iab    64   ab$
      9^Iab    65   ab$
     10^Iab    66   ab$
     11^Iab    67   ab$
     12^Iab    68   ab$
     13^Iab    69   ab$
     14^Iab    70   ab$
     15^Iab    71   ab$
     16^Iab    72   ab$
     17^Iab    73   ab$
?

Best,
наб