Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 1 | #! /bin/sh |
| 2 | |
| 3 | # Script for testing regular expressions with perl to check that PCRE2 handles |
Elliott Hughes | 0c26e19 | 2019-08-07 12:24:46 -0700 | [diff] [blame] | 4 | # them the same. For testing with different versions of Perl, if the first |
| 5 | # argument is -perl then the second is taken as the Perl command to use, and |
| 6 | # both are then removed. If the next argument is "-w", Perl is called with |
| 7 | # "-w", which turns on its warning mode. |
Elliott Hughes | 9bc971b | 2018-07-27 13:23:14 -0700 | [diff] [blame] | 8 | # |
| 9 | # The Perl code has to have "use utf8" and "require Encode" at the start when |
| 10 | # running UTF-8 tests, but *not* for non-utf8 tests. (The "require" would |
| 11 | # actually be OK for non-utf8-tests, but is not always installed, so this way |
| 12 | # the script will always run for these tests.) |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 13 | # |
| 14 | # The desired effect is achieved by making this a shell script that passes the |
Elliott Hughes | 0c26e19 | 2019-08-07 12:24:46 -0700 | [diff] [blame] | 15 | # Perl script to Perl through a pipe. If the next argument is "-utf8", a |
| 16 | # suitable prefix is set up. |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 17 | # |
| 18 | # The remaining arguments, if any, are passed to Perl. They are an input file |
| 19 | # and an output file. If there is one argument, the output is written to |
| 20 | # STDOUT. If Perl receives no arguments, it opens /dev/tty as input, and writes |
| 21 | # output to STDOUT. (I haven't found a way of getting it to use STDIN, because |
| 22 | # of the contorted piping input.) |
| 23 | |
| 24 | perl=perl |
Elliott Hughes | 9bc971b | 2018-07-27 13:23:14 -0700 | [diff] [blame] | 25 | perlarg='' |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 26 | prefix='' |
Elliott Hughes | 9bc971b | 2018-07-27 13:23:14 -0700 | [diff] [blame] | 27 | |
Elliott Hughes | 0c26e19 | 2019-08-07 12:24:46 -0700 | [diff] [blame] | 28 | if [ $# -gt 1 -a "$1" = "-perl" ] ; then |
| 29 | shift |
| 30 | perl=$1 |
| 31 | shift |
| 32 | fi |
| 33 | |
Elliott Hughes | 9bc971b | 2018-07-27 13:23:14 -0700 | [diff] [blame] | 34 | if [ $# -gt 0 -a "$1" = "-w" ] ; then |
| 35 | perlarg="-w" |
| 36 | shift |
| 37 | fi |
| 38 | |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 39 | if [ $# -gt 0 -a "$1" = "-utf8" ] ; then |
| 40 | prefix="use utf8; require Encode;" |
| 41 | shift |
| 42 | fi |
| 43 | |
| 44 | |
| 45 | # The Perl script that follows has a similar specification to pcre2test, and so |
| 46 | # can be given identical input, except that input patterns can be followed only |
| 47 | # by Perl's lower case modifiers and certain other pcre2test modifiers that are |
| 48 | # either handled or ignored: |
| 49 | # |
| 50 | # aftertext interpreted as "print $' afterwards" |
| 51 | # afteralltext ignored |
| 52 | # dupnames ignored (Perl always allows) |
Elliott Hughes | 9bc971b | 2018-07-27 13:23:14 -0700 | [diff] [blame] | 53 | # jitstack ignored |
Elliott Hughes | 653c210 | 2019-01-09 15:41:36 -0800 | [diff] [blame] | 54 | # mark show mark information |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 55 | # no_auto_possess ignored |
Elliott Hughes | 653c210 | 2019-01-09 15:41:36 -0800 | [diff] [blame] | 56 | # no_start_optimize insert (??{""}) at pattern start (disables optimizing) |
| 57 | # -no_start_optimize ignored |
Elliott Hughes | 9bc971b | 2018-07-27 13:23:14 -0700 | [diff] [blame] | 58 | # subject_literal does not process subjects for escapes |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 59 | # ucp sets Perl's /u modifier |
| 60 | # utf invoke UTF-8 functionality |
| 61 | # |
Elliott Hughes | 653c210 | 2019-01-09 15:41:36 -0800 | [diff] [blame] | 62 | # Comment lines are ignored. The #pattern command can be used to set modifiers |
| 63 | # that will be added to each subsequent pattern, after any modifiers it may |
| 64 | # already have. NOTE: this is different to pcre2test where #pattern sets |
| 65 | # defaults which can be overridden on individual patterns. The #subject command |
| 66 | # may be used to set or unset a default "mark" modifier for data lines. This is |
| 67 | # the only use of #subject that is supported. The #perltest, #forbid_utf, and |
| 68 | # #newline_default commands, which are needed in the relevant pcre2test files, |
| 69 | # are ignored. Any other #-command is ignored, with a warning message. |
| 70 | # |
Elliott Hughes | 9bc971b | 2018-07-27 13:23:14 -0700 | [diff] [blame] | 71 | # The data lines must not have any pcre2test modifiers. Unless |
Elliott Hughes | 653c210 | 2019-01-09 15:41:36 -0800 | [diff] [blame] | 72 | # "subject_literal" is on the pattern, data lines are processed as |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 73 | # Perl double-quoted strings, so if they contain " $ or @ characters, these |
| 74 | # have to be escaped. For this reason, all such characters in the |
| 75 | # Perl-compatible testinput1 and testinput4 files are escaped so that they can |
| 76 | # be used for perltest as well as for pcre2test. The output from this script |
| 77 | # should be same as from pcre2test, apart from the initial identifying banner. |
| 78 | # |
| 79 | # The other testinput files are not suitable for feeding to perltest.sh, |
| 80 | # because they make use of the special modifiers that pcre2test uses for |
| 81 | # testing features of PCRE2. Some of these files also contain malformed regular |
| 82 | # expressions, in order to check that PCRE2 diagnoses them correctly. |
| 83 | |
| 84 | (echo "$prefix" ; cat <<'PERLEND' |
| 85 | |
Elliott Hughes | 0c26e19 | 2019-08-07 12:24:46 -0700 | [diff] [blame] | 86 | # The alpha assertions currently give warnings even when -w is not specified. |
| 87 | |
| 88 | no warnings "experimental::alpha_assertions"; |
| 89 | no warnings "experimental::script_run"; |
| 90 | |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 91 | # Function for turning a string into a string of printing chars. |
| 92 | |
| 93 | sub pchars { |
| 94 | my($t) = ""; |
| 95 | if ($utf8) |
| 96 | { |
| 97 | @p = unpack('U*', $_[0]); |
| 98 | foreach $c (@p) |
| 99 | { |
| 100 | if ($c >= 32 && $c < 127) { $t .= chr $c; } |
| 101 | else { $t .= sprintf("\\x{%02x}", $c); |
| 102 | } |
| 103 | } |
| 104 | } |
| 105 | else |
| 106 | { |
| 107 | foreach $c (split(//, $_[0])) |
| 108 | { |
| 109 | if (ord $c >= 32 && ord $c < 127) { $t .= $c; } |
| 110 | else { $t .= sprintf("\\x%02x", ord $c); } |
| 111 | } |
| 112 | } |
| 113 | $t; |
| 114 | } |
| 115 | |
| 116 | |
| 117 | # Read lines from a named file or stdin and write to a named file or stdout; |
| 118 | # lines consist of a regular expression, in delimiters and optionally followed |
| 119 | # by options, followed by a set of test data, terminated by an empty line. |
| 120 | |
| 121 | # Sort out the input and output files |
| 122 | |
| 123 | if (@ARGV > 0) |
| 124 | { |
| 125 | open(INFILE, "<$ARGV[0]") || die "Failed to open $ARGV[0]\n"; |
| 126 | $infile = "INFILE"; |
| 127 | $interact = 0; |
| 128 | } |
| 129 | else |
| 130 | { |
| 131 | open(INFILE, "</dev/tty") || die "Failed to open /dev/tty\n"; |
| 132 | $infile = "INFILE"; |
| 133 | $interact = 1; |
| 134 | } |
| 135 | |
| 136 | if (@ARGV > 1) |
| 137 | { |
| 138 | open(OUTFILE, ">$ARGV[1]") || die "Failed to open $ARGV[1]\n"; |
| 139 | $outfile = "OUTFILE"; |
| 140 | } |
| 141 | else { $outfile = "STDOUT"; } |
| 142 | |
| 143 | printf($outfile "Perl $] Regular Expressions\n\n"); |
| 144 | |
Elliott Hughes | 0c26e19 | 2019-08-07 12:24:46 -0700 | [diff] [blame] | 145 | $extra_modifiers = ""; |
| 146 | $default_show_mark = 0; |
| 147 | |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 148 | # Main loop |
| 149 | |
| 150 | NEXT_RE: |
| 151 | for (;;) |
| 152 | { |
| 153 | printf " re> " if $interact; |
| 154 | last if ! ($_ = <$infile>); |
| 155 | printf $outfile "$_" if ! $interact; |
Elliott Hughes | 653c210 | 2019-01-09 15:41:36 -0800 | [diff] [blame] | 156 | next if ($_ =~ /^\s*$/ || $_ =~ /^#[\s!]/); |
| 157 | |
| 158 | # A few of pcre2test's #-commands are supported, or just ignored. Any others |
| 159 | # cause an error. |
| 160 | |
| 161 | if ($_ =~ /^#pattern(.*)/) |
| 162 | { |
| 163 | $extra_modifiers = $1; |
| 164 | chomp($extra_modifiers); |
| 165 | $extra_modifiers =~ s/\s+$//; |
| 166 | next; |
| 167 | } |
| 168 | elsif ($_ =~ /^#subject(.*)/) |
| 169 | { |
| 170 | $mod = $1; |
| 171 | chomp($mod); |
| 172 | $mod =~ s/\s+$//; |
| 173 | if ($mod =~ s/(-?)mark,?//) |
| 174 | { |
| 175 | $minus = $1; |
| 176 | $default_show_mark = ($minus =~ /^$/); |
| 177 | } |
| 178 | if ($mod !~ /^\s*$/) |
| 179 | { |
| 180 | printf $outfile "** Warning: \"$mod\" in #subject ignored\n"; |
| 181 | } |
| 182 | next; |
| 183 | } |
| 184 | elsif ($_ =~ /^#/) |
| 185 | { |
| 186 | if ($_ !~ /^#newline_default|^#perltest|^#forbid_utf/) |
| 187 | { |
| 188 | printf $outfile "** Warning: #-command ignored: %s", $_; |
| 189 | } |
| 190 | next; |
| 191 | } |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 192 | |
| 193 | $pattern = $_; |
| 194 | |
| 195 | while ($pattern !~ /^\s*(.).*\1/s) |
| 196 | { |
| 197 | printf " > " if $interact; |
| 198 | last if ! ($_ = <$infile>); |
| 199 | printf $outfile "$_" if ! $interact; |
| 200 | $pattern .= $_; |
| 201 | } |
| 202 | |
| 203 | chomp($pattern); |
| 204 | $pattern =~ s/\s+$//; |
| 205 | |
| 206 | # Split the pattern from the modifiers and adjust them as necessary. |
| 207 | |
| 208 | $pattern =~ /^\s*((.).*\2)(.*)$/s; |
| 209 | $pat = $1; |
Elliott Hughes | 653c210 | 2019-01-09 15:41:36 -0800 | [diff] [blame] | 210 | $del = $2; |
| 211 | $mod = "$3,$extra_modifiers"; |
| 212 | $mod =~ s/^,\s*//; |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 213 | |
| 214 | # The private "aftertext" modifier means "print $' afterwards". |
| 215 | |
| 216 | $showrest = ($mod =~ s/aftertext,?//); |
| 217 | |
Elliott Hughes | 9bc971b | 2018-07-27 13:23:14 -0700 | [diff] [blame] | 218 | # The "subject_literal" modifer disables escapes in subjects. |
| 219 | |
| 220 | $subject_literal = ($mod =~ s/subject_literal,?//); |
| 221 | |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 222 | # "allaftertext" is used by pcre2test to print remainders after captures |
| 223 | |
| 224 | $mod =~ s/allaftertext,?//; |
| 225 | |
| 226 | # Detect utf |
| 227 | |
| 228 | $utf8 = $mod =~ s/utf,?//; |
| 229 | |
| 230 | # Remove "dupnames". |
| 231 | |
| 232 | $mod =~ s/dupnames,?//; |
| 233 | |
Elliott Hughes | 9bc971b | 2018-07-27 13:23:14 -0700 | [diff] [blame] | 234 | # Remove "jitstack". |
| 235 | |
| 236 | $mod =~ s/jitstack=\d+,?//; |
| 237 | |
Elliott Hughes | 653c210 | 2019-01-09 15:41:36 -0800 | [diff] [blame] | 238 | # The "mark" modifier requests checking of MARK data */ |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 239 | |
Elliott Hughes | 653c210 | 2019-01-09 15:41:36 -0800 | [diff] [blame] | 240 | $show_mark = $default_show_mark | ($mod =~ s/mark,?//); |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 241 | |
| 242 | # "ucp" asks pcre2test to set PCRE2_UCP; change this to /u for Perl |
| 243 | |
| 244 | $mod =~ s/ucp,?/u/; |
| 245 | |
Elliott Hughes | 653c210 | 2019-01-09 15:41:36 -0800 | [diff] [blame] | 246 | # Remove "no_auto_possess". |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 247 | |
| 248 | $mod =~ s/no_auto_possess,?//; |
Elliott Hughes | 653c210 | 2019-01-09 15:41:36 -0800 | [diff] [blame] | 249 | |
| 250 | # Use no_start_optimize (disable PCRE2 start-up optimization) to disable Perl |
| 251 | # optimization by inserting (??{""}) at the start of the pattern. We may |
| 252 | # also encounter -no_start_optimize from a #pattern setting. |
| 253 | |
| 254 | $mod =~ s/-no_start_optimize,?//; |
| 255 | if ($mod =~ s/no_start_optimize,?//) { $pat =~ s/$del/$del(??{""})/; } |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 256 | |
| 257 | # Add back retained modifiers and check that the pattern is valid. |
| 258 | |
| 259 | $mod =~ s/,//g; |
| 260 | $pattern = "$pat$mod"; |
| 261 | eval "\$_ =~ ${pattern}"; |
| 262 | if ($@) |
| 263 | { |
| 264 | printf $outfile "Error: $@"; |
| 265 | if (! $interact) |
| 266 | { |
| 267 | for (;;) |
| 268 | { |
| 269 | last if ! ($_ = <$infile>); |
| 270 | last if $_ =~ /^\s*$/; |
| 271 | } |
| 272 | } |
| 273 | next NEXT_RE; |
| 274 | } |
| 275 | |
| 276 | # If the /g modifier is present, we want to put a loop round the matching; |
| 277 | # otherwise just a single "if". |
| 278 | |
| 279 | $cmd = ($pattern =~ /g[a-z]*$/)? "while" : "if"; |
| 280 | |
| 281 | # If the pattern is actually the null string, Perl uses the most recently |
| 282 | # executed (and successfully compiled) regex is used instead. This is a |
| 283 | # nasty trap for the unwary! The PCRE2 test suite does contain null strings |
| 284 | # in places - if they are allowed through here all sorts of weird and |
| 285 | # unexpected effects happen. To avoid this, we replace such patterns with |
| 286 | # a non-null pattern that has the same effect. |
| 287 | |
| 288 | $pattern = "/(?#)/$2" if ($pattern =~ /^(.)\1(.*)$/); |
| 289 | |
| 290 | # Read data lines and test them |
| 291 | |
| 292 | for (;;) |
| 293 | { |
| 294 | printf "data> " if $interact; |
| 295 | last NEXT_RE if ! ($_ = <$infile>); |
| 296 | chomp; |
| 297 | printf $outfile "%s", "$_\n" if ! $interact; |
| 298 | |
| 299 | s/\s+$//; # Remove trailing space |
| 300 | s/^\s+//; # Remove leading space |
| 301 | |
| 302 | last if ($_ eq ""); |
| 303 | next if $_ =~ /^\\=(?:\s|$)/; # Comment line |
| 304 | |
Elliott Hughes | 9bc971b | 2018-07-27 13:23:14 -0700 | [diff] [blame] | 305 | if ($subject_literal) |
| 306 | { |
| 307 | $x = $_; |
| 308 | } |
| 309 | else |
| 310 | { |
| 311 | $x = eval "\"$_\""; # To get escapes processed |
| 312 | } |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 313 | |
| 314 | # Empty array for holding results, ensure $REGERROR and $REGMARK are |
| 315 | # unset, then do the matching. |
| 316 | |
| 317 | @subs = (); |
| 318 | |
| 319 | $pushes = "push \@subs,\$&;" . |
| 320 | "push \@subs,\$1;" . |
| 321 | "push \@subs,\$2;" . |
| 322 | "push \@subs,\$3;" . |
| 323 | "push \@subs,\$4;" . |
| 324 | "push \@subs,\$5;" . |
| 325 | "push \@subs,\$6;" . |
| 326 | "push \@subs,\$7;" . |
| 327 | "push \@subs,\$8;" . |
| 328 | "push \@subs,\$9;" . |
| 329 | "push \@subs,\$10;" . |
| 330 | "push \@subs,\$11;" . |
| 331 | "push \@subs,\$12;" . |
| 332 | "push \@subs,\$13;" . |
| 333 | "push \@subs,\$14;" . |
| 334 | "push \@subs,\$15;" . |
| 335 | "push \@subs,\$16;" . |
| 336 | "push \@subs,\$'; }"; |
| 337 | |
| 338 | undef $REGERROR; |
| 339 | undef $REGMARK; |
| 340 | |
| 341 | eval "${cmd} (\$x =~ ${pattern}) {" . $pushes; |
| 342 | |
| 343 | if ($@) |
| 344 | { |
| 345 | printf $outfile "Error: $@\n"; |
| 346 | next NEXT_RE; |
| 347 | } |
| 348 | elsif (scalar(@subs) == 0) |
| 349 | { |
| 350 | printf $outfile "No match"; |
Elliott Hughes | 653c210 | 2019-01-09 15:41:36 -0800 | [diff] [blame] | 351 | if ($show_mark && defined $REGERROR && $REGERROR != 1) |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 352 | { printf $outfile (", mark = %s", &pchars($REGERROR)); } |
| 353 | printf $outfile "\n"; |
| 354 | } |
| 355 | else |
| 356 | { |
| 357 | while (scalar(@subs) != 0) |
| 358 | { |
| 359 | printf $outfile (" 0: %s\n", &pchars($subs[0])); |
| 360 | printf $outfile (" 0+ %s\n", &pchars($subs[17])) if $showrest; |
| 361 | $last_printed = 0; |
| 362 | for ($i = 1; $i <= 16; $i++) |
| 363 | { |
| 364 | if (defined $subs[$i]) |
| 365 | { |
| 366 | while ($last_printed++ < $i-1) |
| 367 | { printf $outfile ("%2d: <unset>\n", $last_printed); } |
| 368 | printf $outfile ("%2d: %s\n", $i, &pchars($subs[$i])); |
| 369 | $last_printed = $i; |
| 370 | } |
| 371 | } |
| 372 | splice(@subs, 0, 18); |
| 373 | } |
| 374 | |
| 375 | # It seems that $REGMARK is not marked as UTF-8 even when use utf8 is |
| 376 | # set and the input pattern was a UTF-8 string. We can, however, force |
| 377 | # it to be so marked. |
| 378 | |
Elliott Hughes | 653c210 | 2019-01-09 15:41:36 -0800 | [diff] [blame] | 379 | if ($show_mark && defined $REGMARK && $REGMARK != 1) |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 380 | { |
| 381 | $xx = $REGMARK; |
| 382 | $xx = Encode::decode_utf8($xx) if $utf8; |
| 383 | printf $outfile ("MK: %s\n", &pchars($xx)); |
| 384 | } |
| 385 | } |
| 386 | } |
| 387 | } |
| 388 | |
Elliott Hughes | 0c26e19 | 2019-08-07 12:24:46 -0700 | [diff] [blame] | 389 | # By closing OUTFILE explicitly, we avoid a Perl warning in -w mode |
| 390 | # "main::OUTFILE" used only once". |
| 391 | |
| 392 | close(OUTFILE) if $outfile eq "OUTFILE"; |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 393 | |
| 394 | PERLEND |
Elliott Hughes | 9bc971b | 2018-07-27 13:23:14 -0700 | [diff] [blame] | 395 | ) | $perl $perlarg - $@ |
Janis Danisevskis | 112c9cc | 2016-03-31 13:35:25 +0100 | [diff] [blame] | 396 | |
| 397 | # End |