Re: coreutils "sort -b -kSTART, END.ENDCHAR" incompatibility with POSIX

[Jim Meyering]

Paul Eggert <address@hidden> wrote:
> Yesterday's email exchange with Stuart Allsop caused me to discover a
> longstanding bug in coreutils "sort".  This bug causes "sort" to be
> incompatible with both POSIX and traditional Unix sort.  Here's a
> proposed patch.
>
> 2004-04-25  Paul Eggert  <address@hidden>
>
>       Fix POSIX-conformance bug: "sort -k 3,3.5b" is supposed to skip
>       leading blanks when computing the location of the field end;
>       it is not supposed to skip trailing blanks.  Solaris 8 "sort"
>       does conform to POSIX.  Also fix the documentation to clarify
>       this and related issues.
>
>       * doc/coreutils.texi (sort invocation): Mention -k earlier, so
>       that the options are in alphabetical order.  Describe how -b works
>       more-accurately; this involves fixing some examples, too.  Mention
>       what happens if the start field falls after an end field or after
>       a line end.  Warn about using -k without -b, -g, -M, -n, or -t.
>       Add an example of how to sort IPv4 addresses and Apache Common
>       Log Format dates.  Remove a duplicate example.
>       (Putting the tools together): Use separate options rather
>       than agglomerating them.
>       * src/sort.c (limfield): Use skipeblanks, not skipsblanks, to
>       decode whether to skip leading blanks.
>       (trailing_blanks): Remove.
>       (fillbuf, getmonth, keycompare): Don't trim trailing blanks.
>
>       * tests/pr/Test.pm: Fix typo in env_default comment.
>       * tests/sort/Test.pm: Likewise.
>       (18c, 18d): Reverse the order of output lines, so that the
>       test cases conform to POSIX.

Nice catch!  Applied.
Thanks for finding and fixing that.