join: new options: --check-order and --nocheck-order.

* src/join.c: Support --check-order and --nocheck-order.
New variables check_input_order, seen_unpairable and
issued_disorder_warning[]. For --check-order, verify that the
input files are in sorted order.  For the default case, check the
order only if there are unpairable lines.
(join): Perform ordering checks after reaching EOF on either
input.
(usage): Mention --check-order and --nocheck-order.
(dupline): Save a copy of the previously-read input line so that
we can detect disorder on the input.
(get_line): Temporarily save a copy of the previous line (by
calling dupline) and check relative ordering (by calling
checkorder) before returning the newly-read line.
(getseq, join): Tell get_line which file we are reading from.
(advance_seq): New function, factoring out some of the code
commonly surrounding calls to getseq.
(checkorder): New function.  Verifies that a pair of consecutive
input lines are in sorted order.
* doc/coreutils.texi (join invocation): Document the new options
--check-order and --nocheck-order.
* tests/join/Test.pm (tv): Added tests for --check-order and
--nocheck-order.
* NEWS: Mention this new feature.

1 files changed, 33 insertions, 2 deletions

diff --git a/tests/join/Test.pm b/tests/join/Test.pm
index 481360489..6d91908f4 100644
--- a/tests/join/Test.pm
+++ b/tests/join/Test.pm
@@ -1,6 +1,6 @@
 # Test "join".
 
-# Copyright (C) 1996, 1999, 2000, 2003, 2004 Free Software Foundation, Inc.
+# Copyright (C) 1996, 1999-2000, 2003-2004, 2008 Free Software Foundation, Inc.
 
 # This program is free software: you can redistribute it and/or modify
 # it under the terms of the GNU General Public License as published by
@@ -140,7 +140,38 @@ my @tv = (
 # FIXME: change this to ensure the diagnostic makes sense
 ['invalid-j', '-j x', {}, "", 1],
 
-);
+# With ordering check, inputs in order
+['chkodr-1', '--check-order',
+  [" a 1\n b 2\n", " a Y\n b Z\n"], "a 1 Y\nb 2 Z\n", 0],
+
+# Without check, inputs in order
+['chkodr-2', '--nocheck-order',
+ [" a 1\n b 2\n", " a Y\n b Z\n"], "a 1 Y\nb 2 Z\n", 0],
+
+# Without check, both inputs out of order (in fact, in reverse order)
+# but all pairable.  Support for this is a GNU extension.
+['chkodr-3', '--nocheck-order',
+ [" b 1\n a 2\n", " b Y\n a Z\n"], "b 1 Y\na 2 Z\n", 0],
+
+# The extension should work without --nocheck-order, since that is the
+# default.
+['chkodr-4', '',
+ [" b 1\n a 2\n", " b Y\n a Z\n"], "b 1 Y\na 2 Z\n", 0],
+
+# With check, both inputs out of order (in fact, in reverse order)
+['chkodr-5', '--check-order',
+ [" b 1\n a 2\n", " b Y\n a Z\n"], "", 1],
+
+# Without order check, both inputs out of order and some lines
+# unpairable.  This is NOT supported by the GNU extension.  All that
+# we really care about for this test is that the return status is
+# zero, since that is the only way to actually verify that the
+# --nocheck-order option had any effect.   We don't actually want to
+# guarantee that join produces this output on stdout.
+['chkodr-6', '--nocheck-order',
+ [" b 1\n a 2\n", " b Y\n c Z\n"], "b 1 Y\n", 0]
+)
+;
 
 
 sub test_vector