Re: ls sort order

[Matthew Schalit]

Hey now,


"Roy C. Sigsbey" wrote:
> 
> Who the FUCK gave you authority to change the sort order for
> the "ls" command.  You have now FUCKED OVER 25 years of progamming.
> The version of "ls" as shipped with RedHat Linux 7.1 is SHIT and
> it's YOUR FAULT!!!


Ouch.  Sound's like Roy's pissed.

Hmmm.  RedHat 7.1?  Don't own it.  But perhaps
I can post the wrong answer, then we'll get the
right one quick like.

 
> ALL files should be listed in STRICT ANSI order, not intermixed
> with the goddamn DOT files!!!


Perhaps.  I only have a Unix box to test on, but it
has a native ls and gnu ls (ver. 4.1).  Both showed the same
search order when I entered 

      ls -al

on my home directory.  Dot files were intermixed with
regular files.  I wonder if ls -al is what Roy's referring to.

Well, if it is, then

    truss ls -al

shows that the files are lxstat'ed in one order:

lxstat(4, ".", 0x0805D63C)                      = 0
lxstat(4, "..", 0x0805D6E0)                     = 0
lxstat(4, ".login", 0x0805D828)                 = 0
lxstat(4, ".profile", 0x0805D8CC)               = 0
lxstat(4, ".vtlrc", 0x0805D970)                 = 0
lxstat(4, ".UpgradeVer7.1.0", 0x0805DA14)       = 0
lxstat(4, "mbox", 0x0805DAB8)                   = 0
lxstat(4, "Java", 0x0805DB5C)                   = 0
lxstat(4, "Dload", 0x0805DC00)                  = 0
lxstat(4, "QuickSco.html", 0x0805DCA4)          = 0
lxstat(4, ".sh_history", 0x0805DD48)            = 0
lxstat(4, "Mail", 0x0805DDEC)                   = 0
lxstat(4, "CpuHog.txt", 0x0805DE90)             = 0
lxstat(4, "Bin", 0x0805DF34)                    = 0
lxstat(4, "Images", 0x0805DFD8)                 = 0
lxstat(4, "mail", 0x0805E07C)                   = 0
lxstat(4, ".dt", 0x0805E120)                    = 0
lxstat(4, ".dtprofile", 0x0805E1C4)             = 0
lxstat(4, ".netscape", 0x0805E268)              = 0
lxstat(4, "nsmail", 0x0805E30C)                 = 0
lxstat(4, ".Xauthority", 0x0805E3B0)            = 0
lxstat(4, "nsnews", 0x0805E454)                 = 0
lxstat(4, "dognome", 0x08060648)                = 0
lxstat(4, ".mime.types", 0x080606EC)            = 0
lxstat(4, ".mailcap", 0x08060790)               = 0
lxstat(4, "upc.bmp", 0x08060834)                = 0
lxstat(4, ".gv", 0x080608D8)                    = 0
getdents(3, 0x0805B240, 2048)                   = 0
close(3)                                        = 0
open("/usr/lib/locale/en_US/LC_COLLATE", O_RDONLY, 01001075150) = 3


But then the files are sorted by the locale defs 
in LC_COLLATE, above, and then displayed in 
semi-alphabetical order:

-rw-------    1 matthew  other           0 Oct 21  1999 .addressbook
-rw-------    1 matthew  other        2285 Oct 21  1999 .addressbook.lu
-rw-------    1 matthew  family        100 Mar 11 12:46 barf
-rw-------    1 matthew  family        170 Feb  6 11:13 .bash_history
drwx------    2 matthew  family       1024 Apr 26 21:41 Bin
-rw-------    1 matthew  family   16777216 May  6 21:36 core.26846
...



So isn't the issue really Roy's default locale?


Then again, there was this:

2001-01-02  Paul Eggert  <address@hidden>

        Make ls -l compatible with POSIX, which requires that the
        behavior of time formats must depend only on LC_TIME, not on
        LC_MESSAGES.


But, that's LC_TIME.


There was this, which could be it:

2000-11-11  Paul Eggert  <address@hidden>

        * src/ls.c (<langinfo.h>): Include if HAVE_LANGINFO_H.
        (DATE_FMT_LANGINFO): New macro, taken from sh-utils/src/date.c.
        (print_long_format): Use it to determine "date" format.
        Do not wrap "%b %e %Y" and "%b %e %H:%M" in _(), as it results in
        incorrect behavior when LC_ALL is unset, LC_TIME is "C", and
        LC_MESSAGES or LANG is set to something disagreeing with the POSIX
        locale.



 
> GO DIRECTLY TO HELL!


There already.



> --
> Roy C. Sigsbey <address@hidden>
> http://www.yourwish.com/


My wish is for binutils ld to be able to find /usr/lib/libc.so.1
and for gcc 3.1 to stop shoving -Qy down gnu-as's ass so 
   ../gcc/configure
won't error out on the first damn test.



Regards,
Matthew