tests/sort/Test.pm

   1 # -*-perl-*-
   2 package Test;
   3 require 5.002;
   4 use strict;
   5
   6 # Tell sort to accept old-style options like `+1'.
   7 $Test::env_default = ['_POSIX2_VERSION=199209'];
   8
   9 my @tv = (
  10 #test   options   input   expected-output   expected-return-code
  11 #
  12 ["n1", '-n', ".01\n0\n", "0\n.01\n", 0],
  13 ["n2", '-n', ".02\n.01\n", ".01\n.02\n", 0],
  14 ["n3", '-n', ".02\n.00\n", ".00\n.02\n", 0],
  15 ["n4", '-n', ".02\n.000\n", ".000\n.02\n", 0],
  16 ["n5", '-n', ".021\n.029\n", ".021\n.029\n", 0],
  17
  18 ["n6", '-n', ".02\n.0*\n", ".0*\n.02\n", 0],
  19 ["n7", '-n', ".02\n.*\n", ".*\n.02\n", 0],
  20 ["n8a", '-s -n -k1,1', ".0a\n.0b\n", ".0a\n.0b\n", 0],
  21 ["n8b", '-s -n -k1,1', ".0b\n.0a\n", ".0b\n.0a\n", 0],
  22 ["n9a", '-s -n -k1,1', ".000a\n.000b\n", ".000a\n.000b\n", 0],
  23 ["n9b", '-s -n -k1,1', ".000b\n.000a\n", ".000b\n.000a\n", 0],
  24 ["n10a", '-s -n -k1,1', ".00a\n.000b\n", ".00a\n.000b\n", 0],
  25 ["n10b", '-s -n -k1,1', ".00b\n.000a\n", ".00b\n.000a\n", 0],
  26 ["n11a", '-s -n -k1,1', ".01a\n.010\n", ".01a\n.010\n", 0],
  27 ["n11b", '-s -n -k1,1', ".010\n.01a\n", ".010\n.01a\n", 0],
  28
  29 ["01a", '', "A\nB\nC\n", "A\nB\nC\n", 0],
  30 #
  31 ["02a", '-c', "A\nB\nC\n", '', 0],
  32 ["02b", '-c', "A\nC\nB\n", '', 1],
  33 ["02c", '-c -k1,1', "a\na b\n", '', 0],
  34 # This should fail because there are duplicate keys
  35 ["02m", '-cu', "A\nA\n", '', 1],
  36 ["02n", '-cu', "A\nB\n", '', 0],
  37 ["02o", '-cu', "A\nB\nB\n", '', 1],
  38 ["02p", '-cu', "B\nA\nB\n", '', 1],
  39 #
  40 ["03a", '-k1', "B\nA\n", "A\nB\n",  0],
  41 ["03b", '-k1,1', "B\nA\n", "A\nB\n",  0],
  42 ["03c", '-k1 -k2', "A b\nA a\n", "A a\nA b\n",  0],
  43 # Fail with a diagnostic when -k specifies field == 0.
  44 ["03d", '-k0', "", "",  2],
  45 # Fail with a diagnostic when -k specifies character == 0.
  46 ["03e", '-k1.0', "", "",  2],
  47 ["03f", '-k1.1,-k0', "", "",  2],
  48 # This is ok.
  49 ["03g", '-k1.1,1.0', "", "",  0],
  50 # This is equivalent to 3f.
  51 ["03h", '-k1.1,1', "", "",  0],
  52 # This too, is equivalent to 3f.
  53 ["03i", '-k1,1', "", "",  0],
  54 #
  55 ["04a", '-nc', "2\n11\n", "",  0],
  56 ["04b", '-n', "11\n2\n", "2\n11\n", 0],
  57 ["04c", '-k1n', "11\n2\n", "2\n11\n", 0],
  58 ["04d", '-k1', "11\n2\n", "11\n2\n", 0],
  59 ["04e", '-k2', "ignored B\nz-ig A\n", "z-ig A\nignored B\n", 0],
  60 #
  61 ["05a", '-k1,2', "A B\nA A\n", "A A\nA B\n", 0],
  62 ["05b", '-k1,2', "A B A\nA A Z\n", "A A Z\nA B A\n", 0],
  63 ["05c", '-k1 -k2', "A B A\nA A Z\n", "A A Z\nA B A\n", 0],
  64 ["05d", '-k2,2', "A B A\nA A Z\n", "A A Z\nA B A\n", 0],
  65 ["05e", '-k2,2', "A B Z\nA A A\n", "A A A\nA B Z\n", 0],
  66 ["05f", '-k2,2', "A B A\nA A Z\n", "A A Z\nA B A\n", 0],
  67 #
  68 ["06a", '-k 1,2', "A B\nA A\n", "A A\nA B\n", 0],
  69 ["06b", '-k 1,2', "A B A\nA A Z\n", "A A Z\nA B A\n", 0],
  70 ["06c", '-k 1 -k 2', "A B A\nA A Z\n", "A A Z\nA B A\n", 0],
  71 ["06d", '-k 2,2', "A B A\nA A Z\n", "A A Z\nA B A\n", 0],
  72 ["06e", '-k 2,2', "A B Z\nA A A\n", "A A A\nA B Z\n", 0],
  73 ["06f", '-k 2,2', "A B A\nA A Z\n", "A A Z\nA B A\n", 0],
  74 #
  75 ["07a", '-k 2,3', "9 a b\n7 a a\n", "7 a a\n9 a b\n", 0],
  76 ["07b", '-k 2,3', "a a b\nz a a\n", "z a a\na a b\n", 0],
  77 ["07c", '-k 2,3', "y k b\nz k a\n", "z k a\ny k b\n", 0],
  78 ["07d", '+1 -3', "y k b\nz k a\n", "z k a\ny k b\n", 0],
  79 #
  80 # report an error for `.' without following char spec
  81 ["08a", '-k 2.,3', "", "", 2],
  82 # report an error for `,' without following POS2
  83 ["08b", '-k 2,', "", "", 2],
  84 #
  85 # Test new -g option.
  86 ["09a", '-g', "1e2\n2e1\n", "2e1\n1e2\n", 0],
  87 # Make sure -n works how we expect.
  88 ["09b", '-n', "1e2\n2e1\n", "1e2\n2e1\n", 0],
  89 ["09c", '-n', "2e1\n1e2\n", "1e2\n2e1\n", 0],
  90 ["09d", '-k2g', "a 1e2\nb 2e1\n", "b 2e1\na 1e2\n", 0],
  91 #
  92 # Bug reported by Roger Peel <R.Peel@ee.surrey.ac.uk>
  93 ["10a", '-t : -k 2.2,2.2', ":ba\n:ab\n", ":ba\n:ab\n", 0],
  94 # Equivalent to above, but using obsolescent `+pos -pos' option syntax.
  95 ["10b", '-t : +1.1 -1.2', ":ba\n:ab\n", ":ba\n:ab\n", 0],
  96 #
  97 # The same as the preceding two, but with input lines reversed.
  98 ["10c", '-t : -k 2.2,2.2', ":ab\n:ba\n", ":ba\n:ab\n", 0],
  99 # Equivalent to above, but using obsolescent `+pos -pos' option syntax.
 100 ["10d", '-t : +1.1 -1.2', ":ab\n:ba\n", ":ba\n:ab\n", 0],
 101 # Try without -t...
 102 # But note that we have to count the delimiting space at the beginning
 103 # of each field that has it.
 104 ["10a0", '-k 2.3,2.3', "z ba\nz ab\n", "z ba\nz ab\n", 0],
 105 ["10a1", '-k 1.2,1.2', "ba\nab\n", "ba\nab\n", 0],
 106 ["10a2", '-b -k 2.2,2.2', "z ba\nz ab\n", "z ba\nz ab\n", 0],
 107 #
 108 # An even simpler example demonstrating the bug.
 109 ["10e", '-k 1.2,1.2', "ab\nba\n", "ba\nab\n", 0],
 110 #
 111 # The way sort works on these inputs (10f and 10g) seems wrong to me.
 112 # See May 30 ChangeLog entry.  POSIX doesn't seem to say one way or
 113 # the other, but that's the way all other sort implementations work.
 114 ["10f", '-t : -k 1.3,1.3', ":ab\n:ba\n", ":ba\n:ab\n", 0],
 115 ["10g", '-k 1.4,1.4', "a ab\nb ba\n", "b ba\na ab\n", 0],
 116 #
 117 # Exercise bug re using -b to skip trailing blanks.
 118 ["11a", '-t: -k1,1b -k2,2', "a\t:a\na :b\n", "a\t:a\na :b\n", 0],
 119 ["11b", '-t: -k1,1b -k2,2', "a :b\na\t:a\n", "a\t:a\na :b\n", 0],
 120 ["11c", '-t: -k2,2b -k3,3', "z:a\t:a\na :b\n", "z:a\t:a\na :b\n", 0],
 121 # Before 1.22m, the first key comparison reported equality.
 122 # With 1.22m, they compare different: "a" sorts before "a\n",
 123 # and the second key spec isn't even used.
 124 ["11d", '-t: -k2,2b -k3,3', "z:a :b\na\t:a\n", "a\t:a\nz:a :b\n", 0],
 125 #
 126 # Exercise bug re comparing `-' and integers.
 127 ["12a", '-n -t: +1', "a:1\nb:-\n", "b:-\na:1\n", 0],
 128 ["12b", '-n -t: +1', "b:-\na:1\n", "b:-\na:1\n", 0],
 129 # Try some other (e.g. `X') invalid character.
 130 ["12c", '-n -t: +1', "a:1\nb:X\n", "b:X\na:1\n", 0],
 131 ["12d", '-n -t: +1', "b:X\na:1\n", "b:X\na:1\n", 0],
 132 # From Karl Heuer
 133 ["13a", '+0.1n', "axx\nb-1\n", "b-1\naxx\n", 0],
 134 ["13b", '+0.1n', "b-1\naxx\n", "b-1\naxx\n", 0],
 135 #
 136 # From Carl Johnson <carlj@cjlinux.home.org>
 137 ["14a", '-d -u', "mal\nmal-\nmala\n", "mal\nmala\n", 0],
 138 # Be sure to fix the (translate && ignore) case in keycompare.
 139 ["14b", '-f -d -u', "mal\nmal-\nmala\n", "mal\nmala\n", 0],
 140 #
 141 # Experiment with -i.
 142 ["15a", '-i -u', "a\na\1\n", "a\n", 0],
 143 ["15b", '-i -u', "a\n\1a\n", "a\n", 0],
 144 ["15c", '-i -u', "a\1\na\n", "a\1\n", 0],
 145 ["15d", '-i -u', "\1a\na\n", "\1a\n", 0],
 146 ["15e", '-i -u', "a\n\1\1\1\1\1a\1\1\1\1\n", "a\n", 0],
 147
 148 # From Erick Branderhorst -- fixed around 1.19e
 149 ["16a", '-f',
 150  "éminence\nüberhaupt\n's-Gravenhage\naëroclub\nAag\naagtappels\n",
 151  "'s-Gravenhage\nAag\naagtappels\naëroclub\néminence\nüberhaupt\n",
 152  0],
 153
 154 # This provokes a one-byte memory overrun of a malloc'd block for versions
 155 # of sort from textutils-1.19p and before.
 156 ["17", '-c', "xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx\n", "", 0],
 157
 158 # POSIX says -n no longer implies -b, so here we're comparing ` 9' and `10'.
 159 ["18a", '-k1.1,1.2n', " 901\n100\n", " 901\n100\n", 0],
 160
 161 # Just like above, because the global `-b' has no effect on the
 162 # key specifier when a key-specific option (`n' in this case) is used.
 163 ["18b", '-b -k1.1,1.2n', " 901\n100\n", " 901\n100\n", 0],
 164
 165 # Here we're comparing ` 90' and `10', because the `b' on the key-end specifier
 166 # makes sort ignore leading blanks when determining that key's *end*.
 167 ["18c", '-k1.1,1.2nb', " 901\n100\n", "100\n 901\n", 0],
 168
 169 # Here we're comparing `9' and `10', because the `b' on the key-start specifier
 170 # makes sort ignore leading blanks when determining that key's *start*.
 171 ["18d", '-k1.1b,1.2n', " 901\n100\n", " 901\n100\n", 0],
 172
 173 # This compares `90' and `10', as it ignores leading blanks for both
 174 # key start and key end.
 175 ["18e", '-nb -k1.1,1.2', " 901\n100\n", "100\n 901\n", 0],
 176
 177 # This looks odd, but works properly -- 2nd keyspec is never
 178 # used because all lines are different.
 179 ["19a", '+0 +1nr', "b 2\nb 1\nb 3\n", "b 1\nb 2\nb 3\n", 0],
 180
 181 # The test *intended* by the author of the above, but using the
 182 # more-intuitive POSIX-style -k options.
 183 ["19b", '-k1,1 -k2nr', "b 2\nb 1\nb 3\n", "b 3\nb 2\nb 1\n", 0],
 184
 185 # This test failed when sort-1.22 was compiled on a Next x86 system
 186 # without optimization.  Without optimization gcc uses the buggy version
 187 # of memcmp in the Next C library.  With optimization, gcc uses its
 188 # (working) builtin version.  Test case form William Lewis.
 189 ["20a", '',
 190  "_________U__free\n_________U__malloc\n_________U__abort\n_________U__memcpy\n_________U__memset\n_________U_dyld_stub_binding_helper\n_________U__malloc\n_________U___iob\n_________U__abort\n_________U__fprintf\n",
 191  "_________U___iob\n_________U__abort\n_________U__abort\n_________U__fprintf\n_________U__free\n_________U__malloc\n_________U__malloc\n_________U__memcpy\n_________U__memset\n_________U_dyld_stub_binding_helper\n",
 192  0],
 193
 194 # Demonstrate that folding changes the ordering of e.g. A, a, and _
 195 # because while they normally (in the C locale) collate like A, _, a,
 196 # when using -f, `a' is compared as if it were `A'.
 197 ["21a", '',    "A\na\n_\n", "A\n_\na\n", 0],
 198 ["21b", '-f',  "A\na\n_\n", "A\na\n_\n", 0],
 199 ["21c", '-f',  "a\nA\n_\n", "A\na\n_\n", 0],
 200 ["21d", '-f',  "_\na\nA\n", "A\na\n_\n", 0],
 201 ["21e", '-f',  "a\n_\nA\n", "A\na\n_\n", 0],
 202 ["21f", '-fs', "A\na\n_\n", "A\na\n_\n", 0],
 203 ["21g", '-fu', "a\n_\n", "a\n_\n", 0],
 204
 205 # This test failed until 1.22f.  From Zvi Har'El.
 206 ["22a", '-k 2,2fd -k 1,1r', "3 b\n4 B\n", "4 B\n3 b\n", 0],
 207 ["22b", '-k 2,2d  -k 1,1r', "3 b\n4 b\n", "4 b\n3 b\n", 0],
 208
 209 ["no-file1", 'no-file', {}, '', 2],
 210 # This test failed until 1.22f.  Sort didn't give an error.
 211 # From Will Edgington.
 212 ["o-no-file1", '-o no-such-file no-such-file', {}, '', 2],
 213
 214 ["create-empty", '-o no/such/file /dev/null', {}, '', 2],
 215
 216 # From Paul Eggert.  This was fixed in textutils-1.22k.
 217 ["neg-nls", '-n', "-1\n-9\n", "-9\n-1\n", 0],
 218
 219 # From Paul Eggert.  This was fixed in textutils-1.22m.
 220 # The bug was visible only when using the internationalized sorting code
 221 # (i.e., not when configured with --disable-nls).
 222 ["nul-nls", '', "\0b\n\0a\n", "\0a\n\0b\n", 0],
 223
 224 # Paul Eggert wrote:
 225 # I tested the revised `sort' against Solaris `sort', and found a
 226 # discrepancy that turns out to be a longstanding bug in GNU sort.
 227 # POSIX.2 specifies that a newline is part of the input line, and should
 228 # be significant during comparison; but with GNU sort the newline is
 229 # insignificant.  Here is an example of the bug:
 230 #
 231 #       $ od -c t
 232 #       0000000  \n  \t  \n
 233 #       0000003
 234 #       $ sort t | od -c
 235 #       0000000  \n  \t  \n
 236 #       0000003
 237 #
 238 # The correct output of the latter command should be
 239 #
 240 #       0000000  \t  \n  \n
 241 #       0000003
 242 #
 243 # because \t comes before \n in the collating sequence, and the trailing
 244 # \n's are part of the input line.
 245 ["use-nl", '', "\n\t\n", "\n\t\n", 0],
 246
 247 # Specifying two -o options should evoke a failure
 248 ["o2", '-o x -o y', '', '', 2],
 249
 250 # -t '\0' is accepted, as of coreutils-5.0.91
 251 ['nul-tab', "-k2,2 -t '\\0'", "a\0z\01\nb\0y\02\n", "b\0y\02\na\0z\01\n", 0],
 252 );
 253
 254 sub test_vector
 255 {
 256   return @tv;
 257 }
 258
 259 1;