On Thu, Apr 16, 2009 at 8:40 PM, Don Cragun <dcragun at sonic.net> wrote: >> Date: Thu, 16 Apr 2009 12:11:42 -0500 >> From: Jason King <jason at ansipunx.net> >> >> On Thu, Apr 16, 2009 at 11:49 AM, Glenn Skinner <glenn.skinner at sun.com> >> wrote: >>> ? ?Date: Wed, 15 Apr 2009 14:05:29 -0700 >>> ? ?From: "Garrett D'Amore" <gdamore at sun.com> > ?... ... ... >>> >>> But before I give my +1, a quick question: ?Have you verified that >>> adding the new flags to the /usr/xpg*/bin/ls variants is ok from the >>> standpoint of standards conformance? ?If the answer is affirmative, >>> I'm happy with the proposal. >> >> Unfortunately I'm not a standards language lawyer, so it's difficult >> for me to know what things I need to look out for. >> >> The one thing I'm aware of is the cmdline option processing (and to >> not reorder arguments). ?I was told using getopt_long_only(3c) avoids >> this issue (which is what the code does). > > POSIX standards and the SUS specifications allow implementations to > provide additional options. > > According to the CLIP case (PSARC/1999/645), the following are required: > 1. ?Every long option must have a short (single character) equivalent, and > 2. ?optional option arguments are not allowed. > I don't see short option equivalents for --block-size, --si, --color (with > optional > option-argument), --file-type, and --time-style (with optional > option-argument). > The ARC can allow exemptions for violations, but the exemptions have > to be requested and the violations need to be noted in the man page.
This becomes interesting -- there aren't a lot of letters left. The options without equivalent short options match what gnu ls does. That of course doesn't mean we can't add equivalent short options (assuming an appropriate letter can be found). For color at least, -G might work -- it'd make it compatible with the BSD color ls, however GNU ls -G is equivalent to Solaris ls -o (according to the GNU ls manpage). I'm not sure what the correct balance should be. The big feature is of course colorized output (as well as the -k option), the others were mostly just incredibly trivial to implement, and can easily be ripped out if deemed of little value. I don't think 100% compatibility is possible (at least without morphing the case), so mostly just 'what bits do we want, what do we not'. > > Also, there still is no man page in the case directory (at least not visible > outside of Sun), but the one-pager shows -w=NN as a new flag. ?There > should not be an '=' separating a short option from its option-argument. Hmm.. I'll ping Garrett and see if he can place it into the directory (it's my first attempt at such a thing, so I would expect it might need weaking). the -w=NN was just bad on my part, -w NN should be the correct format. That was a typo on my part..
