[ChangeLog]

When decoding, always allow newlines in input, with almost no
performance impact.
* src/base64.c (do_decode): Initialize decode context.
Call base64_decode one more time, after all input is processed.
(usage): When decoding, newlines are always accepted.

* tests/misc/base64: Add a bunch of tests, for the above.
* gl/lib/base64.c: Include <string.h>.
(base64_decode_ctx_init, get_4, decode_4): New functions.
(base64_decode): Efficiently handle interspersed newlines.
(base64_decode_alloc): Update signature.
* gl/lib/base64.h (struct base64_decode_context): Define.
(base64_decode_ctx_init): Add prototype.
(base64_decode, base64_decode_alloc): Update prototypes.
[doc/ChangeLog]
* coreutils.texi (base64 invocation): When decoding, newlines
are always accepted.

1 files changed, 88 insertions, 16 deletions

diff --git a/tests/misc/base64 b/tests/misc/base64
index 60332a044..70295fb7f 100755
--- a/tests/misc/base64
+++ b/tests/misc/base64
@@ -2,7 +2,7 @@
 # -*- perl -*-
 # Exercise base64.
 
-# Copyright (C) 2006 Free Software Foundation, Inc.
+# Copyright (C) 2006, 2007 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
@@ -37,21 +37,69 @@ use strict;
 # Turn off localisation of executable's ouput.
 @ENV{qw(LANGUAGE LANG LC_ALL)} = ('C') x 3;
 
-(my $a39 = 'YWFh' x 13) =~ s/(.{5})/$1\n/g;
+# Return the encoding of a string of N 'a's.
+sub enc($)
+{
+  my ($n) = @_;
+  my %remainder = ( 0 => '', 1 => 'YQ==', 2 => 'YWE=' );
+  return 'YWFh' x ($n / 3) . $remainder{$n % 3};
+}
+
+# Construct an encoded string of length 4KB, using 3K "a"s.
+my $a3k = enc 3072;
+my @a3k_nl;
+# A few copies, each with different number of newlines at the start.
+for my $k (0..3)
+  {
+    (my $t = $a3k) =~ s/^/"\n"x $k/e;
+    push @a3k_nl, $t;
+  }
+
+# Return a copy of S, with newlines inserted every WIDTH bytes.
+# Ensure that the result (if not the empty string) is newline-terminated.
+sub wrap($$)
+{
+  my ($s, $width) = @_;
+  $s =~ s/(.{$width})/$1\n/g;
+  substr ($s, -1, 1) ne "\n"
+    and $s .= "\n";
+  return $s;
+}
 
 my @Tests =
     (
      ['empty', {IN=>''}, {OUT=>""}],
      ['inout', {IN=>'a'}, {OUT=>"YQ==\n"}],
      ['wrap', '--wrap 0', {IN=>'foo'}, {OUT=>'Zm9v'}],
-     ['wrap5-39', '--wrap=5', {IN=>'a' x 39}, {OUT=>"${a39}\n"}],
-     ['wrap5-40', '--wrap=5', {IN=>'a' x 40}, {OUT=>"${a39}YQ=\n=\n"}],
-     ['wrap5-41', '--wrap=5', {IN=>'a' x 41}, {OUT=>"${a39}YWE\n=\n"}],
-     ['wrap5-42', '--wrap=5', {IN=>'a' x 42}, {OUT=>"${a39}YWF\nh\n"}],
-     ['wrap5-43', '--wrap=5', {IN=>'a' x 43}, {OUT=>"${a39}YWF\nhYQ==\n"}],
-     ['wrap5-44', '--wrap=5', {IN=>'a' x 44}, {OUT=>"${a39}YWF\nhYWE=\n"}],
-     ['wrap5-45', '--wrap=5', {IN=>'a' x 45}, {OUT=>"${a39}YWF\nhYWFh\n"}],
-     ['wrap5-46', '--wrap=5', {IN=>'a' x 46}, {OUT=>"${a39}YWF\nhYWFh\nYQ==\n"}],
+     ['wrap5-39', '--wrap=5', {IN=>'a' x 39}, {OUT=>wrap enc(39),5}],
+     ['wrap5-40', '--wrap=5', {IN=>'a' x 40}, {OUT=>wrap enc(40),5}],
+     ['wrap5-41', '--wrap=5', {IN=>'a' x 41}, {OUT=>wrap enc(41),5}],
+     ['wrap5-42', '--wrap=5', {IN=>'a' x 42}, {OUT=>wrap enc(42),5}],
+     ['wrap5-43', '--wrap=5', {IN=>'a' x 43}, {OUT=>wrap enc(43),5}],
+     ['wrap5-44', '--wrap=5', {IN=>'a' x 44}, {OUT=>wrap enc(44),5}],
+     ['wrap5-45', '--wrap=5', {IN=>'a' x 45}, {OUT=>wrap enc(45),5}],
+     ['wrap5-46', '--wrap=5', {IN=>'a' x 46}, {OUT=>wrap enc(46),5}],
+
+     ['buf-1',   '--decode', {IN=>enc 1}, {OUT=>'a' x 1}],
+     ['buf-2',   '--decode', {IN=>enc 2}, {OUT=>'a' x 2}],
+     ['buf-3',   '--decode', {IN=>enc 3}, {OUT=>'a' x 3}],
+     ['buf-4',   '--decode', {IN=>enc 4}, {OUT=>'a' x 4}],
+     # 4KB worth of input.
+     ['buf-4k0', '--decode', {IN=>enc 3072+0}, {OUT=>'a' x (3072+0)}],
+     ['buf-4k1', '--decode', {IN=>enc 3072+1}, {OUT=>'a' x (3072+1)}],
+     ['buf-4k2', '--decode', {IN=>enc 3072+2}, {OUT=>'a' x (3072+2)}],
+     ['buf-4k3', '--decode', {IN=>enc 3072+3}, {OUT=>'a' x (3072+3)}],
+     ['buf-4km1','--decode', {IN=>enc 3072-1}, {OUT=>'a' x (3072-1)}],
+     ['buf-4km2','--decode', {IN=>enc 3072-2}, {OUT=>'a' x (3072-2)}],
+     ['buf-4km3','--decode', {IN=>enc 3072-3}, {OUT=>'a' x (3072-3)}],
+     ['buf-4km4','--decode', {IN=>enc 3072-4}, {OUT=>'a' x (3072-4)}],
+
+     # Exercise the case in which the final base-64 byte is
+     # in a buffer all by itself.
+     ['b4k-1',   '--decode', {IN=>$a3k_nl[1]}, {OUT=>'a' x (3072+0)}],
+     ['b4k-2',   '--decode', {IN=>$a3k_nl[2]}, {OUT=>'a' x (3072+0)}],
+     ['b4k-3',   '--decode', {IN=>$a3k_nl[3]}, {OUT=>'a' x (3072+0)}],
+
      ['baddecode', '--decode', {IN=>'a'}, {OUT=>""},
       {ERR_SUBST => 's/.*: invalid input//'}, {ERR => "\n"}, {EXIT => 1}],
      ['baddecode2', '--decode', {IN=>'ab'}, {OUT=>"i"},
@@ -65,13 +113,18 @@ my @Tests =
     );
 
 # For each non-failing test, create a --decode test using the
-# expected output (with newlines removed) as input.
+# expected output as input.  Also, add tests inserting newlines.
 my @new;
 foreach my $t (@Tests)
   {
     my $exit_val;
     my $in;
-    my $out;
+    my @out;
+
+    # If the test has a single option of "--decode", then skip it.
+    !ref $t->[1] && $t->[1] eq '--decode'
+      and next;
+
     foreach my $e (@$t)
       {
         ref $e && ref $e eq 'HASH'
@@ -80,11 +133,30 @@ foreach my $t (@Tests)
 	  and $exit_val = $e->{EXIT};
 	defined $e->{IN}
 	  and $in = $e->{IN};
-	defined $e->{OUT}
-	  and ($out = $e->{OUT}) =~ tr/\n//d;
+	if (defined $e->{OUT})
+	  {
+	    my $t = $e->{OUT};
+	    push @out, $t;
+	    my $len = length $t;
+	    foreach my $i (0..$len)
+	      {
+		my $u = $t;
+		substr ($u, $i, 0) = "\n";
+		push @out, $u;
+		10 <= $i
+		  and last;
+	      }
+	  }
+      }
+    $exit_val
+      and next;
+
+    my $i = 0;
+    foreach my $o (@out)
+      {
+	push @new, ["d$i-$t->[0]", '--decode', {IN => $o}, {OUT => $in}];
+	++$i;
       }
-    defined $out && ! $exit_val
-      and push @new, ["d-$t->[0]", '--decode', {IN => $out}, {OUT => $in}];
   }
 push @Tests, @new;