[PATCH] sort -m: don't segfault when output file is also an input file

[Jim Meyering]

Thanks to Otavio Salvador for finding/reporting this.

Here's the patch I'm considering:

>From 570beb56f58bb087a614af885bec7e9cf6b19423 Mon Sep 17 00:00:00 2001
From: Jim Meyering <address@hidden>
Date: Wed, 22 Apr 2009 08:45:27 +0200
Subject: [PATCH] sort -m: don't segfault when output file is also an input file

* src/sort.c (avoid_trashing_input): Fix an off-by-one error and
guard the use of memmove.
* NEWS (Bug fixes): Mention it.
* tests/misc/sort: Add tests to exercise the offending code.
* THANKS: Update.
Reported by Otavio Salvador in http://bugs.debian.org/525048.
---
 NEWS            |    6 +++++-
 THANKS          |    1 +
 src/sort.c      |   10 ++++++----
 tests/misc/sort |   17 ++++++++++++++++-
 4 files changed, 28 insertions(+), 6 deletions(-)

diff --git a/NEWS b/NEWS
index 5951bb5..ee44a23 100644
--- a/NEWS
+++ b/NEWS
@@ -1,9 +1,13 @@
 GNU coreutils NEWS                                    -*- outline -*-

-* Noteworthy changes in release ?.? (????-??-??) [?]
+* Noteworthy changes in release 7.3 (????-??-??) [?]

 ** Bug fixes

+  sort -m no longer segfaults when its output file is also an input file.
+  E.g., with this, touch 1; sort -m -o 1 1, sort would segfault.
+  [introduced in coreutils-7.2]
+
   ls now aligns output correctly in the presence of abbreviated month
   names from the locale database that have differing widths.

diff --git a/THANKS b/THANKS
index fe523fe..876a6b6 100644
--- a/THANKS
+++ b/THANKS
@@ -435,6 +435,7 @@ Ole Laursen                         address@hidden
 Oliver Kiddle                       address@hidden
 Ørn E. Hansen                       address@hidden
 Oskar Liljeblad                     address@hidden
+Otavio Salvador                     address@hidden
 Pádraig Brady                       address@hidden
 Patrick Mauritz                     address@hidden
 Paul D. Smith                       address@hidden
diff --git a/src/sort.c b/src/sort.c
index 2e6ce87..f48d727 100644
--- a/src/sort.c
+++ b/src/sort.c
@@ -1,5 +1,5 @@
 /* sort - sort lines of text (with all kinds of options).
-   Copyright (C) 1988, 1991-2008 Free Software Foundation, Inc.
+   Copyright (C) 1988, 1991-2009 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
@@ -2602,18 +2602,20 @@ avoid_trashing_input (struct sortfile *files, size_t 
ntemps,
          pid_t pid;
          char *temp = create_temp (&tftp, &pid);
          size_t num_merged = 0;
-         while (i + num_merged < nfiles)
+         do
            {
              num_merged += mergefiles (&files[i], 0, nfiles - i, tftp, temp);
              files[i].name = temp;
              files[i].pid = pid;

-             memmove(&files[i], &files[i + num_merged],
-                     num_merged * sizeof *files);
+             if (i + num_merged < nfiles)
+               memmove(&files[i + 1], &files[i + num_merged],
+                       num_merged * sizeof *files);
              ntemps += 1;
              nfiles -= num_merged - 1;;
              i += num_merged;
            }
+         while (i < nfiles);
        }
     }

diff --git a/tests/misc/sort b/tests/misc/sort
index 4f377df..a27f47a 100755
--- a/tests/misc/sort
+++ b/tests/misc/sort
@@ -1,6 +1,6 @@
 #!/usr/bin/perl

-# Copyright (C) 2008 Free Software Foundation, Inc.
+# Copyright (C) 2008, 2009 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
@@ -339,6 +339,21 @@ my @Tests =
 ["sort-numeric", '--sort=numeric', {IN=>".01\n0\n"}, {OUT=>"0\n.01\n"}],
 ["sort-gennum", '--sort=general-numeric',
   {IN=>"1e2\n2e1\n"}, {OUT=>"2e1\n1e2\n"}],
+
+# -m with output file also used as an input file
+# In coreutils-7.2, this caused a segfault.
+# This test looks a little strange.  Here's why:
+# since we're using "-o f", standard output will be empty, hence OUT=>''
+# We still want to ensure that the output file, "f" has expected contents,
+# hence the added CMP=> directive.
+["output-is-input", '-m -o f', {IN=> {f=> "a\n"}}, {OUT=>''},
+ {CMP=> ["a\n", {'f'=> undef}]} ],
+["output-is-input-2", '-m -o f', {OUT=>''},
+ {IN=> {f=> "a\n"}}, {IN=> {g=> "b\n"}}, {IN=> {h=> "c\n"}},
+ {CMP=> ["a\nb\nc\n", {'f'=> undef}]} ],
+["output-is-input-3", '-m -o f', {OUT=>''},
+ {IN=> {g=> "a\n"}}, {IN=> {h=> "b\n"}}, {IN=> {f=> "c\n"}},
+ {CMP=> ["a\nb\nc\n", {'f'=> undef}]} ],
 );

 # Add _POSIX2_VERSION=199209 to the environment of each test
--
1.6.3.rc1.205.g37f8