From: Stuart Henderson Subject: Re: mtime format in ls -l To: Ingo Schwarze , tech@openbsd.org, Jan Stary Cc: Crystal Kolipe Date: Fri, 16 Jan 2026 20:45:18 +0000 On 2026/01/16 19:59, Ingo Schwarze wrote: > > I think this is a case where being less specific is more valuable: > being less specific conveys a feeling to the reader that the output > is mainly intended for human consumption. Also, being very specific > has little value as simply running ls(1) in an arbitrary directory > makes it clear how the output looks (roughly, in usual cases). > > So, here is what i propose; OK? > Ingo > > > Index: ls.1 > =================================================================== > RCS file: /cvs/src/bin/ls/ls.1,v > diff -u -r1.79 ls.1 > --- ls.1 24 Oct 2016 13:46:58 -0000 1.79 > +++ ls.1 16 Jan 2026 18:47:59 -0000 > @@ -281,8 +281,7 @@ > .Fl g ) , > group, > size in bytes, > -time of last modification > -.Pq Dq mmm dd HH:MM , > +date and time or date and year of last modification, > and the pathname. > In addition, for each directory whose contents are displayed, the first > line displayed is the total number of blocks used by the files in the > Good idea - ok. On 2026/01/16 20:04, Crystal Kolipe wrote: > Any half sensible script that was parsing dates from ls(1) would surely be > using -T together with -l, (or -g, or -n), anyway, so the 'mmm dd HH:MM' would > not even apply in that case. > > I wonder if the 'see also' section should include a reference to stat(1), > since ls(1) doesn't provide an option equivalent to -P in df(1), for obvious > consumption by scripts. I don't think there's any good portable tool to do this from the shell. GNU ls doesn't do -T and we don't have the overengineered --time-style, and stat(1) is probably the least portable of all the standard unix utilities I've come across. Seems like a job for perl.