Fast and extensible multi-platform HTTP/1-2-3 web server with automatic HTTPS
Mirrored from https://github.com/Slazee/caddy.git
Synced Nov 11, 01:00 UTC
| 1 | // Copyright 2015 Matthew Holt and The Caddy Authors |
| 2 | // |
| 3 | // Licensed under the Apache License, Version 2.0 (the "License"); |
| 4 | // you may not use this file except in compliance with the License. |
| 5 | // You may obtain a copy of the License at |
| 6 | // |
| 7 | // http://www.apache.org/licenses/LICENSE-2.0 |
| 8 | // |
| 9 | // Unless required by applicable law or agreed to in writing, software |
| 10 | // distributed under the License is distributed on an "AS IS" BASIS, |
| 11 | // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 12 | // See the License for the specific language governing permissions and |
| 13 | // limitations under the License. |
| 14 | |
| 15 | package caddyfile |
| 16 | |
| 17 | import ( |
| 18 | "bytes" |
| 19 | "os" |
| 20 | "reflect" |
| 21 | "strings" |
| 22 | "testing" |
| 23 | ) |
| 24 | |
| 25 | func TestMain(m *testing.M) { |
| 26 | // Pin an empty environment so replaceEnvVars is deterministic for fuzzing. |
| 27 | os.Clearenv() |
| 28 | os.Exit(m.Run()) |
| 29 | } |
| 30 | |
| 31 | func TestFormatWindowsTrailingBackslash(t *testing.T) { |
| 32 | input := "localhost {\nroot * C:\\site\\\nfile_server\n}\n" |
| 33 | want := "localhost {\n\troot * C:\\site\\\n\tfile_server\n}\n" |
| 34 | for _, in := range []string{input, strings.ReplaceAll(input, "\n", "\r\n")} { |
| 35 | out := Format([]byte(in)) |
| 36 | if strings.ReplaceAll(string(out), "\r\n", "\n") != want { |
| 37 | t.Fatalf("got %q, want %q", out, want) |
| 38 | } |
| 39 | if !bytes.Equal(Format(out), out) { |
| 40 | t.Fatal("formatting is not idempotent") |
| 41 | } |
| 42 | blocks, err := Parse("", out) |
| 43 | if err != nil { |
| 44 | t.Fatal(err) |
| 45 | } |
| 46 | if len(blocks) != 1 || len(blocks[0].Segments) != 2 { |
| 47 | t.Fatalf("expected separate root and file_server directives, got %+v", blocks) |
| 48 | } |
| 49 | } |
| 50 | } |
| 51 | |
| 52 | func TestFormatter(t *testing.T) { |
| 53 | for i, tc := range []struct { |
| 54 | description string |
| 55 | input string |
| 56 | expect string |
| 57 | }{ |
| 58 | { |
| 59 | description: "very simple", |
| 60 | input: `abc def |
| 61 | g hi jkl |
| 62 | mn`, |
| 63 | expect: `abc def |
| 64 | g hi jkl |
| 65 | mn`, |
| 66 | }, |
| 67 | { |
| 68 | description: "basic indentation, line breaks, and nesting", |
| 69 | input: ` a |
| 70 | b |
| 71 | |
| 72 | c { |
| 73 | d |
| 74 | } |
| 75 | |
| 76 | e { f |
| 77 | } |
| 78 | |
| 79 | |
| 80 | |
| 81 | g { |
| 82 | h { |
| 83 | i |
| 84 | } |
| 85 | } |
| 86 | |
| 87 | j { k { |
| 88 | l |
| 89 | } |
| 90 | } |
| 91 | |
| 92 | m { |
| 93 | n { o |
| 94 | } |
| 95 | p { q r |
| 96 | s } |
| 97 | } |
| 98 | |
| 99 | { |
| 100 | { t |
| 101 | u |
| 102 | |
| 103 | v |
| 104 | |
| 105 | w |
| 106 | } |
| 107 | }`, |
| 108 | expect: `a |
| 109 | b |
| 110 | |
| 111 | c { |
| 112 | d |
| 113 | } |
| 114 | |
| 115 | e { |
| 116 | f |
| 117 | } |
| 118 | |
| 119 | g { |
| 120 | h { |
| 121 | i |
| 122 | } |
| 123 | } |
| 124 | |
| 125 | j { |
| 126 | k { |
| 127 | l |
| 128 | } |
| 129 | } |
| 130 | |
| 131 | m { |
| 132 | n { |
| 133 | o |
| 134 | } |
| 135 | p { |
| 136 | q r |
| 137 | s |
| 138 | } |
| 139 | } |
| 140 | |
| 141 | { |
| 142 | { |
| 143 | t |
| 144 | u |
| 145 | |
| 146 | v |
| 147 | |
| 148 | w |
| 149 | } |
| 150 | }`, |
| 151 | }, |
| 152 | { |
| 153 | description: "block spacing", |
| 154 | input: `a{ |
| 155 | b |
| 156 | } |
| 157 | |
| 158 | c{ d |
| 159 | }`, |
| 160 | expect: `a{ |
| 161 | b |
| 162 | } |
| 163 | |
| 164 | c{ d |
| 165 | }`, |
| 166 | }, |
| 167 | { |
| 168 | // sanctioned divergence: invalid glued braces left literal (design decision) |
| 169 | description: "advanced spacing", |
| 170 | input: `abc { |
| 171 | def |
| 172 | }ghi{ |
| 173 | jkl mno |
| 174 | pqr}`, |
| 175 | expect: `abc { |
| 176 | def |
| 177 | }ghi{ |
| 178 | jkl mno |
| 179 | pqr}`, |
| 180 | }, |
| 181 | { |
| 182 | description: "env var placeholders", |
| 183 | input: `{$A} |
| 184 | |
| 185 | b { |
| 186 | {$C} |
| 187 | } |
| 188 | |
| 189 | d { {$E} |
| 190 | } |
| 191 | |
| 192 | { {$F} |
| 193 | } |
| 194 | `, |
| 195 | expect: `{$A} |
| 196 | |
| 197 | b { |
| 198 | {$C} |
| 199 | } |
| 200 | |
| 201 | d { |
| 202 | {$E} |
| 203 | } |
| 204 | |
| 205 | { |
| 206 | {$F} |
| 207 | }`, |
| 208 | }, |
| 209 | { |
| 210 | description: "env var placeholders with port", |
| 211 | input: `:{$PORT}`, |
| 212 | expect: `:{$PORT}`, |
| 213 | }, |
| 214 | { |
| 215 | description: "comments", |
| 216 | input: `#a "\n" |
| 217 | |
| 218 | #b { |
| 219 | c |
| 220 | } |
| 221 | |
| 222 | d { |
| 223 | e#f |
| 224 | # g |
| 225 | } |
| 226 | |
| 227 | h { # i |
| 228 | }`, |
| 229 | expect: `#a "\n" |
| 230 | |
| 231 | #b { |
| 232 | c |
| 233 | } |
| 234 | |
| 235 | d { |
| 236 | e#f |
| 237 | # g |
| 238 | } |
| 239 | |
| 240 | h { # i |
| 241 | }`, |
| 242 | }, |
| 243 | { |
| 244 | description: "quotes and escaping", |
| 245 | input: `"a \"b\" "#c |
| 246 | d |
| 247 | |
| 248 | e { |
| 249 | "f" |
| 250 | } |
| 251 | |
| 252 | g { "h" |
| 253 | } |
| 254 | |
| 255 | i { |
| 256 | "foo |
| 257 | bar" |
| 258 | } |
| 259 | |
| 260 | j { |
| 261 | "\"k\" l m" |
| 262 | }`, |
| 263 | expect: `"a \"b\" "#c |
| 264 | d |
| 265 | |
| 266 | e { |
| 267 | "f" |
| 268 | } |
| 269 | |
| 270 | g { |
| 271 | "h" |
| 272 | } |
| 273 | |
| 274 | i { |
| 275 | "foo |
| 276 | bar" |
| 277 | } |
| 278 | |
| 279 | j { |
| 280 | "\"k\" l m" |
| 281 | }`, |
| 282 | }, |
| 283 | { |
| 284 | description: "bad nesting (too many open)", |
| 285 | input: `a |
| 286 | { |
| 287 | { |
| 288 | }`, |
| 289 | expect: `a { |
| 290 | { |
| 291 | } |
| 292 | `, |
| 293 | }, |
| 294 | { |
| 295 | // sanctioned divergence: invalid glued braces left literal (design decision) |
| 296 | description: "bad nesting (too many close)", |
| 297 | input: `a |
| 298 | { |
| 299 | { |
| 300 | }}}`, |
| 301 | expect: `a { |
| 302 | { |
| 303 | }}} |
| 304 | `, |
| 305 | }, |
| 306 | { |
| 307 | description: "json", |
| 308 | input: `foo |
| 309 | bar "{\"key\":34}" |
| 310 | `, |
| 311 | expect: `foo |
| 312 | bar "{\"key\":34}"`, |
| 313 | }, |
| 314 | { |
| 315 | description: "escaping after spaces", |
| 316 | input: `foo \"literal\"`, |
| 317 | expect: `foo \"literal\"`, |
| 318 | }, |
| 319 | { |
| 320 | description: "simple placeholders as standalone tokens", |
| 321 | input: `foo {bar}`, |
| 322 | expect: `foo {bar}`, |
| 323 | }, |
| 324 | { |
| 325 | description: "simple placeholders within tokens", |
| 326 | input: `foo{bar} foo{bar}baz`, |
| 327 | expect: `foo{bar} foo{bar}baz`, |
| 328 | }, |
| 329 | { |
| 330 | // sanctioned divergence: invalid glued braces left literal (design decision) |
| 331 | description: "placeholders and malformed braces", |
| 332 | input: `foo{bar} foo{ bar}baz`, |
| 333 | expect: `foo{bar} foo{ bar}baz`, |
| 334 | }, |
| 335 | { |
| 336 | description: "hash within string is not a comment", |
| 337 | input: `redir / /some/#/path`, |
| 338 | expect: `redir / /some/#/path`, |
| 339 | }, |
| 340 | { |
| 341 | description: "brace does not fold into comment above", |
| 342 | input: `# comment |
| 343 | { |
| 344 | foo |
| 345 | }`, |
| 346 | expect: `# comment |
| 347 | { |
| 348 | foo |
| 349 | }`, |
| 350 | }, |
| 351 | { |
| 352 | description: "matthewpi/vscode-caddyfile-support#13", |
| 353 | input: `{ |
| 354 | email {$ACMEEMAIL} |
| 355 | #debug |
| 356 | } |
| 357 | |
| 358 | block { |
| 359 | } |
| 360 | `, |
| 361 | expect: `{ |
| 362 | email {$ACMEEMAIL} |
| 363 | #debug |
| 364 | } |
| 365 | |
| 366 | block { |
| 367 | } |
| 368 | `, |
| 369 | }, |
| 370 | { |
| 371 | description: "matthewpi/vscode-caddyfile-support#13 - bad formatting", |
| 372 | input: `{ |
| 373 | email {$ACMEEMAIL} |
| 374 | #debug |
| 375 | } |
| 376 | |
| 377 | block { |
| 378 | } |
| 379 | `, |
| 380 | expect: `{ |
| 381 | email {$ACMEEMAIL} |
| 382 | #debug |
| 383 | } |
| 384 | |
| 385 | block { |
| 386 | } |
| 387 | `, |
| 388 | }, |
| 389 | { |
| 390 | description: "keep heredoc as-is", |
| 391 | input: `block { |
| 392 | heredoc <<HEREDOC |
| 393 | Here's more than one space Here's more than one space |
| 394 | HEREDOC |
| 395 | } |
| 396 | `, |
| 397 | expect: `block { |
| 398 | heredoc <<HEREDOC |
| 399 | Here's more than one space Here's more than one space |
| 400 | HEREDOC |
| 401 | } |
| 402 | `, |
| 403 | }, |
| 404 | { |
| 405 | description: "Mixing heredoc with regular part", |
| 406 | input: `block { |
| 407 | heredoc <<HEREDOC |
| 408 | Here's more than one space Here's more than one space |
| 409 | HEREDOC |
| 410 | respond "More than one space will be eaten" 200 |
| 411 | } |
| 412 | |
| 413 | block2 { |
| 414 | heredoc <<HEREDOC |
| 415 | Here's more than one space Here's more than one space |
| 416 | HEREDOC |
| 417 | respond "More than one space will be eaten" 200 |
| 418 | } |
| 419 | `, |
| 420 | expect: `block { |
| 421 | heredoc <<HEREDOC |
| 422 | Here's more than one space Here's more than one space |
| 423 | HEREDOC |
| 424 | respond "More than one space will be eaten" 200 |
| 425 | } |
| 426 | |
| 427 | block2 { |
| 428 | heredoc <<HEREDOC |
| 429 | Here's more than one space Here's more than one space |
| 430 | HEREDOC |
| 431 | respond "More than one space will be eaten" 200 |
| 432 | } |
| 433 | `, |
| 434 | }, |
| 435 | { |
| 436 | description: "Heredoc as regular token", |
| 437 | input: `block { |
| 438 | heredoc <<HEREDOC "More than one space will be eaten" |
| 439 | } |
| 440 | `, |
| 441 | expect: `block { |
| 442 | heredoc <<HEREDOC "More than one space will be eaten" |
| 443 | } |
| 444 | `, |
| 445 | }, |
| 446 | { |
| 447 | description: "Escape heredoc", |
| 448 | input: `block { |
| 449 | heredoc \<<HEREDOC |
| 450 | respond "More than one space will be eaten" 200 |
| 451 | } |
| 452 | `, |
| 453 | expect: `block { |
| 454 | heredoc \<<HEREDOC |
| 455 | respond "More than one space will be eaten" 200 |
| 456 | } |
| 457 | `, |
| 458 | }, |
| 459 | { |
| 460 | description: "Preserve braces wrapped by backquotes", |
| 461 | input: "block {respond `All braces should remain: {{now | date \"2006\"}}`}", |
| 462 | expect: "block {respond `All braces should remain: {{now | date \"2006\"}}`}", |
| 463 | }, |
| 464 | { |
| 465 | description: "Preserve braces wrapped by quotes", |
| 466 | input: "block {respond \"All braces should remain: {{now | date `2006`}}\"}", |
| 467 | expect: "block {respond \"All braces should remain: {{now | date `2006`}}\"}", |
| 468 | }, |
| 469 | { |
| 470 | description: "Preserve quoted brace arguments", |
| 471 | input: "block {\n\trespond \"{\"\n\trespond \"}\"\n}", |
| 472 | expect: "block {\n\trespond \"{\"\n\trespond \"}\"\n}", |
| 473 | }, |
| 474 | { |
| 475 | description: "Preserve quoted backticks and backticked quotes", |
| 476 | input: "block { respond \"`\" } block { respond `\"`}", |
| 477 | expect: "block {\n\trespond \"`\"\n}\n\nblock {\n\trespond `\"`\n}", |
| 478 | }, |
| 479 | { |
| 480 | description: "No trailing space on line before env variable", |
| 481 | input: `{ |
| 482 | a |
| 483 | |
| 484 | {$ENV_VAR} |
| 485 | } |
| 486 | `, |
| 487 | expect: `{ |
| 488 | a |
| 489 | |
| 490 | {$ENV_VAR} |
| 491 | } |
| 492 | `, |
| 493 | }, |
| 494 | { |
| 495 | description: "issue #7425: multiline backticked string indentation", |
| 496 | input: `https://localhost:8953 { |
| 497 | respond ` + "`" + `Here are some random numbers: |
| 498 | |
| 499 | {{randNumeric 16}} |
| 500 | |
| 501 | Hope this helps.` + "`" + ` |
| 502 | }`, |
| 503 | expect: "https://localhost:8953 {\n\trespond `Here are some random numbers:\n\n{{randNumeric 16}}\n\nHope this helps.`\n}", |
| 504 | }, |
| 505 | { |
| 506 | description: "imports before global options block keep standalone brace", |
| 507 | input: `import ./conf.d/matcher_my_subnet.caddy |
| 508 | import ./conf.d/matcher_not_my_subnet.caddy |
| 509 | { |
| 510 | order crowdsec first |
| 511 | order appsec after crowdsec |
| 512 | }`, |
| 513 | expect: `import ./conf.d/matcher_my_subnet.caddy |
| 514 | import ./conf.d/matcher_not_my_subnet.caddy |
| 515 | { |
| 516 | order crowdsec first |
| 517 | order appsec after crowdsec |
| 518 | }`, |
| 519 | }, |
| 520 | { |
| 521 | description: "keep an opening brace that ends the input", |
| 522 | input: `localhost {`, |
| 523 | expect: `localhost {`, |
| 524 | }, |
| 525 | { |
| 526 | description: "keep a standalone opening brace that ends the input", |
| 527 | input: `{`, |
| 528 | expect: `{`, |
| 529 | }, |
| 530 | { |
| 531 | description: "keep adjacent opening braces that end the input", |
| 532 | input: `{{`, |
| 533 | // Glued braces are a literal token, not structural blocks. |
| 534 | expect: `{{`, |
| 535 | }, |
| 536 | { |
| 537 | description: "keep spaced opening braces that end the input", |
| 538 | input: `{ {`, |
| 539 | expect: `{ |
| 540 | {`, |
| 541 | }, |
| 542 | } { |
| 543 | // the formatter should output a trailing newline, |
| 544 | // even if the tests aren't written to expect that |
| 545 | if !strings.HasSuffix(tc.expect, "\n") { |
| 546 | tc.expect += "\n" |
| 547 | } |
| 548 | |
| 549 | actual := Format([]byte(tc.input)) |
| 550 | |
| 551 | if string(actual) != tc.expect { |
| 552 | t.Errorf("\n[TEST %d: %s]\n====== EXPECTED ======\n%s\n====== ACTUAL ======\n%s^^^^^^^^^^^^^^^^^^^^^", |
| 553 | i, tc.description, string(tc.expect), string(actual)) |
| 554 | } |
| 555 | } |
| 556 | } |
| 557 | |
| 558 | func TestFormatImportStandaloneBrace(t *testing.T) { |
| 559 | // immediately after a top-level import line: keep the brace standalone |
| 560 | in1 := "import a.caddy\nimport b.caddy\n{\n\torder x first\n}\n" |
| 561 | if got := string(Format([]byte(in1))); got != in1 { |
| 562 | t.Errorf("standalone brace after import not preserved:\n got %q\nwant %q", got, in1) |
| 563 | } |
| 564 | // An intervening blank line must not turn the global options block into an |
| 565 | // import block mapping. |
| 566 | in2 := "import a.caddy\n\n{\n\torder x first\n}\n" |
| 567 | want2 := in2 |
| 568 | if got := string(Format([]byte(in2))); got != want2 { |
| 569 | t.Errorf("blank-line case:\n got %q\nwant %q", got, want2) |
| 570 | } |
| 571 | } |
| 572 | |
| 573 | func TestFormatDeepNestingCapsIndent(t *testing.T) { |
| 574 | // Brace matching remains unbounded, but rendered indentation is capped to |
| 575 | // keep linear-size malicious input from producing quadratic output. |
| 576 | var b strings.Builder |
| 577 | const depth = 1000 |
| 578 | for range depth { |
| 579 | b.WriteString("l {\n") |
| 580 | } |
| 581 | b.WriteString("x\n") |
| 582 | for range depth { |
| 583 | b.WriteString("}\n") |
| 584 | } |
| 585 | out := string(Format([]byte(b.String()))) |
| 586 | if !strings.Contains(out, "\n"+strings.Repeat("\t", maxIndent)+"x\n") { |
| 587 | t.Errorf("expected %d-tab indent for deepest token", maxIndent) |
| 588 | } |
| 589 | if len(out) > b.Len()+(2*depth+1)*maxIndent { |
| 590 | t.Fatalf("formatted output grew unexpectedly: input=%d output=%d", b.Len(), len(out)) |
| 591 | } |
| 592 | } |
| 593 | |
| 594 | func TestFormatPreservesLiteralBraceArguments(t *testing.T) { |
| 595 | for _, tc := range []struct { |
| 596 | in string |
| 597 | want string |
| 598 | }{ |
| 599 | {"localhost\nrespond foo{\n", "localhost\nrespond foo{\n"}, |
| 600 | {"localhost\nrespond {} 200\n", "localhost\nrespond {} 200\n"}, |
| 601 | } { |
| 602 | before, err := Tokenize([]byte(tc.in), "") |
| 603 | if err != nil { |
| 604 | t.Fatalf("input %q: %v", tc.in, err) |
| 605 | } |
| 606 | out := Format([]byte(tc.in)) |
| 607 | if string(out) != tc.want { |
| 608 | t.Errorf("Format(%q) = %q, want %q", tc.in, out, tc.want) |
| 609 | } |
| 610 | after, err := Tokenize(out, "") |
| 611 | if err != nil { |
| 612 | t.Fatalf("formatted input %q no longer parses: %v\nout=%q", tc.in, err, out) |
| 613 | } |
| 614 | if !sameTokenTexts(before, after) { |
| 615 | t.Errorf("formatting changed parser tokens for %q:\nbefore=%v\nafter=%v\nout=%q", tc.in, before, after, out) |
| 616 | } |
| 617 | } |
| 618 | } |
| 619 | |
| 620 | func TestFormatEscapedDelimitersDoesNotDisableFormatting(t *testing.T) { |
| 621 | in := "localhost {\nrespond foo\\\"bar\nrespond foo\\`bar\n}\n" |
| 622 | want := "localhost {\n\trespond foo\\\"bar\n\trespond foo\\`bar\n}\n" |
| 623 | if got := string(Format([]byte(in))); got != want { |
| 624 | t.Errorf("got %q, want %q", got, want) |
| 625 | } |
| 626 | } |
| 627 | |
| 628 | func TestFinalizeHeredocLargeBody(t *testing.T) { |
| 629 | line := strings.Repeat("x", 1024) |
| 630 | var val strings.Builder |
| 631 | for range 1024 { |
| 632 | val.WriteByte('\t') |
| 633 | val.WriteString(line) |
| 634 | val.WriteByte('\n') |
| 635 | } |
| 636 | val.WriteString("\tEND") |
| 637 | |
| 638 | got, err := new(lexer).finalizeHeredoc([]rune(val.String()), "END") |
| 639 | if err != nil { |
| 640 | t.Fatal(err) |
| 641 | } |
| 642 | wantLen := 1024*(len(line)+1) - 1 |
| 643 | if len(string(got)) != wantLen { |
| 644 | t.Fatalf("large heredoc length = %d, want %d", len(string(got)), wantLen) |
| 645 | } |
| 646 | } |
| 647 | |
| 648 | func TestWrapUnbracedSiteRejectsLeadingImport(t *testing.T) { |
| 649 | in := "import common.caddy\nlocalhost\nrespond 200\n" |
| 650 | want := string(Format([]byte(in))) |
| 651 | if got := string(FormatWithOptions([]byte(in), FormatOptions{WrapUnbracedSite: true})); got != want { |
| 652 | t.Errorf("leading import was wrapped:\n got %q\nwant %q", got, want) |
| 653 | } |
| 654 | } |
| 655 | |
| 656 | func TestWrapUnbracedSiteAfterNestedDirective(t *testing.T) { |
| 657 | in := "localhost\nroute {\nrespond ok\n}\nfile_server\n" |
| 658 | want := "localhost {\n\troute {\n\t\trespond ok\n\t}\n\tfile_server\n}\n" |
| 659 | if got := string(FormatWithOptions([]byte(in), FormatOptions{WrapUnbracedSite: true})); got != want { |
| 660 | t.Errorf("got %q, want %q", got, want) |
| 661 | } |
| 662 | } |
| 663 | |
| 664 | func TestFormatAngleNotQuirked(t *testing.T) { |
| 665 | if got := string(Format([]byte("foo < bar"))); got != "foo < bar\n" { |
| 666 | t.Errorf("got %q, want %q", got, "foo < bar\n") |
| 667 | } |
| 668 | } |
| 669 | |
| 670 | func TestFormatEmptyBlocksExpand(t *testing.T) { |
| 671 | cases := []struct{ in, want string }{ |
| 672 | {"route {}", "route {}\n"}, |
| 673 | {"route { }", "route {\n}\n"}, |
| 674 | {"a { b {} }", "a {\n\tb {}\n}\n"}, |
| 675 | } |
| 676 | for _, c := range cases { |
| 677 | if got := string(Format([]byte(c.in))); got != c.want { |
| 678 | t.Errorf("in %q: got %q, want %q", c.in, got, c.want) |
| 679 | } |
| 680 | } |
| 681 | } |
| 682 | |
| 683 | func TestFormatTokenAfterCloseBraceBreaks(t *testing.T) { |
| 684 | in := "a {\n\tb {\n\t\tc\n\t} d\n}" |
| 685 | want := "a {\n\tb {\n\t\tc\n\t}\n\td\n}\n" |
| 686 | if got := string(Format([]byte(in))); got != want { |
| 687 | t.Errorf("got %q, want %q", got, want) |
| 688 | } |
| 689 | } |
| 690 | |
| 691 | func TestFormatCommentsOnBraceLines(t *testing.T) { |
| 692 | cases := []struct{ in, want string }{ |
| 693 | {"site {\n\tfoo\n} # after close\n", "site {\n\tfoo\n} # after close\n"}, |
| 694 | {"site { # note\n\tfoo\n}\n", "site { # note\n\tfoo\n}\n"}, |
| 695 | {"site # note\n{\n\tfoo\n}\n", "site { # note\n\tfoo\n}\n"}, |
| 696 | } |
| 697 | for _, c := range cases { |
| 698 | if got := string(Format([]byte(c.in))); got != c.want { |
| 699 | t.Errorf("in %q:\n got %q\nwant %q", c.in, got, c.want) |
| 700 | } |
| 701 | } |
| 702 | } |
| 703 | |
| 704 | func TestFormatBlankLineCapAfterComment(t *testing.T) { |
| 705 | got := string(Format([]byte("foo # inline\n\n\nbar\n"))) |
| 706 | want := "foo # inline\n\nbar\n" |
| 707 | if got != want { |
| 708 | t.Errorf("got %q, want %q", got, want) |
| 709 | } |
| 710 | } |
| 711 | |
| 712 | func TestFormatContinuationHangingIndent(t *testing.T) { |
| 713 | in := "route {\n\treverse_proxy \\\n\ta \\\n\tb\n}\n" |
| 714 | want := "route {\n\treverse_proxy \\\n\t\ta \\\n\t\tb\n}\n" |
| 715 | if got := string(Format([]byte(in))); got != want { |
| 716 | t.Errorf("got %q, want %q", got, want) |
| 717 | } |
| 718 | } |
| 719 | |
| 720 | func TestFormatStrayCloseBraceKeepsTokens(t *testing.T) { |
| 721 | for _, in := range []string{"0 }", "a }"} { |
| 722 | once := Format([]byte(in)) |
| 723 | twice := Format(once) |
| 724 | if !bytes.Equal(once, twice) { |
| 725 | t.Errorf("not idempotent for %q:\n once=%q\ntwice=%q", in, once, twice) |
| 726 | } |
| 727 | a, err1 := Parse("Caddyfile", []byte(in)) |
| 728 | if err1 != nil { |
| 729 | t.Fatalf("input %q did not parse: %v", in, err1) |
| 730 | } |
| 731 | b, err2 := Parse("Caddyfile", once) |
| 732 | if err2 != nil { |
| 733 | t.Fatalf("formatted %q did not parse: %v", in, err2) |
| 734 | } |
| 735 | if !sameStructure(a, b) { |
| 736 | t.Errorf("stray close brace merged tokens for %q: got %q", in, once) |
| 737 | } |
| 738 | } |
| 739 | } |
| 740 | |
| 741 | func TestFormatIncompleteHeredocIdempotent(t *testing.T) { |
| 742 | for _, in := range []string{"0 <<0", "x <<E"} { |
| 743 | once := Format([]byte(in)) |
| 744 | twice := Format(once) |
| 745 | if !bytes.Equal(once, twice) { |
| 746 | t.Errorf("not idempotent for %q:\n once=%q\ntwice=%q", in, once, twice) |
| 747 | } |
| 748 | } |
| 749 | } |
| 750 | |
| 751 | // formatTokenText returns the format-mode non-comment token-text sequence for |
| 752 | // input, or nil and false on a lex error. It is the basis of the semantic- |
| 753 | // preservation invariant: Format must reproduce this exact sequence. Comment |
| 754 | // tokens are omitted because the formatter intentionally normalizes comment |
| 755 | // whitespace (e.g. trimming a trailing space, "# " -> "#"), which is not a |
| 756 | // semantic change. |
| 757 | func formatTokenText(in []byte) ([]string, bool) { |
| 758 | toks, err := Lex(in, "", LexOptions{Comments: true, Raw: true}) |
| 759 | if err != nil { |
| 760 | return nil, false |
| 761 | } |
| 762 | var texts []string |
| 763 | for _, t := range toks { |
| 764 | if !t.isComment { |
| 765 | texts = append(texts, t.Text) |
| 766 | } |
| 767 | } |
| 768 | return texts, true |
| 769 | } |
| 770 | |
| 771 | // formatAndParseLexersDisagree reports whether the format-mode lexer and the |
| 772 | // parse-path lexer produce different token-text sequences for the input (the |
| 773 | // parse baseline is newline-terminated to match Format's mandatory trailing |
| 774 | // newline, and comments — which the parse path discards — are ignored on the |
| 775 | // format side). When the two lexers already disagree before any formatting, the |
| 776 | // input is degenerate or structurally ambiguous — an unterminated or lone |
| 777 | // quote/backtick, an escaped quote swallowing to EOF, a "{}" or a "{"/"}" glued |
| 778 | // to a literal word that only one path splits — and has no well-defined |
| 779 | // formatting. Such inputs are a sanctioned divergence excluded from the |
| 780 | // invariant. Inputs on which both lexers agree, including the stray-close-brace |
| 781 | // bug's "0 }" and standalone braces in argument position, stay in scope. |
| 782 | func formatAndParseLexersDisagree(in []byte) bool { |
| 783 | fmtToks, err := Lex(in, "", LexOptions{Comments: true, Raw: true}) |
| 784 | if err != nil { |
| 785 | return true |
| 786 | } |
| 787 | var fmtText []string |
| 788 | for _, t := range fmtToks { |
| 789 | if t.isComment { |
| 790 | continue |
| 791 | } |
| 792 | fmtText = append(fmtText, t.Text) |
| 793 | // A non-quoted token whose verbatim source ends in whitespace, contains an |
| 794 | // escaped quote, or whose text carries a newline results from an escaped |
| 795 | // quote ("\"") swallowing trailing whitespace to EOF. The format-mode and |
| 796 | // parse-path lexers can agree on such a token, but Format emits the source |
| 797 | // verbatim and then trims trailing whitespace, changing the token on |
| 798 | // re-lex. Treat these degenerate escaped-quote tokens as a disagreement. |
| 799 | if t.wasQuoted == 0 { |
| 800 | r := t.Raw() |
| 801 | if n := len(r); n > 0 && (r[n-1] == ' ' || r[n-1] == '\t' || r[n-1] == '\v' || r[n-1] == '\f' || r[n-1] == '\r' || r[n-1] == '\n') { |
| 802 | return true |
| 803 | } |
| 804 | if strings.Contains(r, `\"`) || strings.Contains(r, "\\`") { |
| 805 | return true |
| 806 | } |
| 807 | if strings.ContainsRune(t.Text, '\n') { |
| 808 | return true |
| 809 | } |
| 810 | } |
| 811 | } |
| 812 | base := in |
| 813 | if n := len(base); n == 0 || base[n-1] != '\n' { |
| 814 | base = append(append([]byte{}, base...), '\n') |
| 815 | } |
| 816 | parseToks, err := Tokenize(base, "") |
| 817 | if err != nil { |
| 818 | return true |
| 819 | } |
| 820 | parseText := make([]string, len(parseToks)) |
| 821 | for i, t := range parseToks { |
| 822 | parseText[i] = t.Text |
| 823 | } |
| 824 | return !reflect.DeepEqual(fmtText, parseText) |
| 825 | } |
| 826 | |
| 827 | // hasHeredocOpenerShapedToken reports whether the format-mode lexer produces a |
| 828 | // token whose text begins with "<<" (a heredoc opener shape). Such a token is a |
| 829 | // literal word only because no newline follows it (or a separating space breaks |
| 830 | // the heredoc), but Format's mandatory trailing newline can turn it into a |
| 831 | // heredoc opener, so the formatted output re-lexes to a different token stream. |
| 832 | // This degenerate opener is a sanctioned divergence excluded from the invariant; |
| 833 | // valid heredocs (whose token text does not begin with "<<") stay in scope. It |
| 834 | // is detected by token text, so a "<<" split by a stripped carriage return |
| 835 | // ("<\r<") is caught too. |
| 836 | func hasHeredocOpenerShapedToken(in []byte) bool { |
| 837 | toks, err := Lex(in, "", LexOptions{Comments: true, Raw: true}) |
| 838 | if err != nil { |
| 839 | return true |
| 840 | } |
| 841 | for _, t := range toks { |
| 842 | if t.wasQuoted == 0 && !t.isComment && strings.HasPrefix(t.Text, "<<") { |
| 843 | return true |
| 844 | } |
| 845 | } |
| 846 | return false |
| 847 | } |
| 848 | |
| 849 | // TestFormatDoesNotMutateInput guards the fallback paths, which trim the input |
| 850 | // and append the mandatory newline. bytes.TrimSpace returns a subslice sharing |
| 851 | // the caller's backing array, so appending to it without copying writes into |
| 852 | // the caller's buffer. os.ReadFile returns a slice with spare capacity, so this |
| 853 | // corrupted the input that "caddy fmt --diff" compares its output against. |
| 854 | func TestFormatDoesNotMutateInput(t *testing.T) { |
| 855 | cases := []string{ |
| 856 | "foo \"unterminated \n\n", // unterminated quote + trailing whitespace |
| 857 | "a b\\ \n\n", // literal trailing backslash + whitespace |
| 858 | ` \ \`, // fallback with significant leading whitespace |
| 859 | " site {\n\tfoo\n} \n\n", // ordinary input with surrounding whitespace |
| 860 | } |
| 861 | for _, in := range cases { |
| 862 | // Spare capacity, as os.ReadFile and pooled buffers provide. |
| 863 | buf := append(make([]byte, 0, len(in)+8), in...) |
| 864 | out := Format(buf) |
| 865 | if string(buf) != in { |
| 866 | t.Errorf("Format mutated its input:\n in %q\n after %q", in, string(buf)) |
| 867 | } |
| 868 | if len(out) > 0 && len(buf) > 0 && &out[0] == &buf[0] { |
| 869 | t.Errorf("Format output aliases its input for %q", in) |
| 870 | } |
| 871 | } |
| 872 | } |
| 873 | |
| 874 | func FuzzFormatIdempotent(f *testing.F) { |
| 875 | for _, s := range []string{ |
| 876 | "", " ", "a{\nb\n}", "site {\n\tfoo # c\n}\n", "x <<E\nhi\nE\n", |
| 877 | // Tokens that swallow end-of-input, and a lone CR that does not. |
| 878 | "\"\"\"", "\"\"`", "``\\\"", "0 0\r", |
| 879 | `0 \`, ` \ \`, // dangling continuation and significant leading space |
| 880 | } { |
| 881 | f.Add([]byte(s)) |
| 882 | } |
| 883 | f.Fuzz(func(t *testing.T, in []byte) { |
| 884 | once := Format(in) |
| 885 | twice := Format(once) |
| 886 | if !bytes.Equal(once, twice) { |
| 887 | t.Errorf("not idempotent:\n once=%q\ntwice=%q", once, twice) |
| 888 | } |
| 889 | }) |
| 890 | } |
| 891 | |
| 892 | func FuzzFormatNoPanic(f *testing.F) { |
| 893 | f.Add([]byte("\x00\x00")) |
| 894 | f.Add([]byte("`unterminated")) |
| 895 | f.Add([]byte("x <<E\nno end marker")) |
| 896 | f.Fuzz(func(t *testing.T, in []byte) { |
| 897 | _ = Format(in) // must not panic |
| 898 | }) |
| 899 | } |
| 900 | |
| 901 | func FuzzFormatSemanticPreserve(f *testing.F) { |
| 902 | for _, s := range []string{"site {\n\troot * /srv\n\tfile_server\n}\n"} { |
| 903 | f.Add([]byte(s)) |
| 904 | } |
| 905 | f.Fuzz(func(t *testing.T, in []byte) { |
| 906 | // The semantic-preservation invariant: formatting only changes structural |
| 907 | // whitespace, so the format-mode token-text sequence of Format(in) must |
| 908 | // equal that of the input. This is stronger and more robust than comparing |
| 909 | // parse structures, which the parse path's lenient/quirky grouping of |
| 910 | // ambiguous brace layouts (e.g. standalone braces in argument position) |
| 911 | // would spuriously flag even though every token is preserved. |
| 912 | // |
| 913 | // One class is genuinely out of scope: a heredoc-opener-shaped token |
| 914 | // ("<<MARKER") that stays a literal word only because no newline follows |
| 915 | // it. Format's mandatory trailing newline turns it into a real heredoc |
| 916 | // opener, so the output legitimately re-lexes to a different token stream. |
| 917 | // Valid heredocs (whose token text does not begin with "<<") stay in scope. |
| 918 | // Also out of scope: any input on which the format-mode and parse-path |
| 919 | // lexers already disagree (unterminated/lone quotes, escaped quotes at EOF, |
| 920 | // "{}" or a brace glued to a literal word) — these are degenerate or |
| 921 | // structurally ambiguous and have no well-defined formatting. |
| 922 | if hasHeredocOpenerShapedToken(in) || formatAndParseLexersDisagree(in) { |
| 923 | return |
| 924 | } |
| 925 | // Compare the format-mode token text of a newline-terminated baseline |
| 926 | // against that of Format's output. Format always ends its output in a |
| 927 | // single newline (Invariant); normalizing the baseline the same way keeps |
| 928 | // the comparison about token content, not the mandatory trailing newline. |
| 929 | base := in |
| 930 | if n := len(base); n == 0 || base[n-1] != '\n' { |
| 931 | base = append(append([]byte{}, base...), '\n') |
| 932 | } |
| 933 | before, ok := formatTokenText(base) |
| 934 | if !ok { |
| 935 | return // input does not lex; Format falls back and there is nothing to preserve |
| 936 | } |
| 937 | out := Format(in) |
| 938 | after, ok := formatTokenText(out) |
| 939 | if !ok { |
| 940 | t.Fatalf("formatted output no longer lexes\ninput=%q\nout=%q", in, out) |
| 941 | } |
| 942 | if !reflect.DeepEqual(before, after) { |
| 943 | t.Errorf("Format changed the token stream\ninput=%q\nout=%q\nbefore=%q\nafter=%q", in, out, before, after) |
| 944 | } |
| 945 | }) |
| 946 | } |
| 947 | |
| 948 | // TestFormatFuzzerAngles covers tricky / pathological inputs that a fuzzer would |
| 949 | // exercise. Every case asserts idempotency (Format(Format(x)) == Format(x)) and |
| 950 | // that Format does not panic. The subset of clearly-idiomatic inputs also asserts |
| 951 | // an exact expected output. |
| 952 | func TestFormatFuzzerAngles(t *testing.T) { |
| 953 | type tc struct { |
| 954 | name string |
| 955 | input string |
| 956 | exactExpect string // non-empty → assert this exact output in addition to idempotency |
| 957 | } |
| 958 | |
| 959 | cases := []tc{ |
| 960 | // ---- clearly-idiomatic: assert exact output ---- |
| 961 | |
| 962 | { |
| 963 | // Empty input: Format always emits a single trailing newline. |
| 964 | name: "empty input", |
| 965 | input: "", |
| 966 | exactExpect: "\n", |
| 967 | }, |
| 968 | { |
| 969 | // Whitespace-only input collapses to the same single newline. |
| 970 | name: "whitespace-only input", |
| 971 | input: " \n\t\n", |
| 972 | exactExpect: "\n", |
| 973 | }, |
| 974 | { |
| 975 | // A bare env-var placeholder is a literal token and must be preserved. |
| 976 | name: "env var {$X}", |
| 977 | input: "{$X}", |
| 978 | exactExpect: "{$X}\n", |
| 979 | }, |
| 980 | { |
| 981 | // Default-value env var. |
| 982 | name: "env var {$X:def}", |
| 983 | input: "{$X:def}", |
| 984 | exactExpect: "{$X:def}\n", |
| 985 | }, |
| 986 | { |
| 987 | // Minimal env var with only the sigil. |
| 988 | name: "env var {$}", |
| 989 | input: "{$}", |
| 990 | exactExpect: "{$}\n", |
| 991 | }, |
| 992 | { |
| 993 | // Backtick token as the first token inside a block must indent correctly. |
| 994 | name: "backtick as first token after {", |
| 995 | input: "a {\n\t`foo`\n}\n", |
| 996 | exactExpect: "a {\n\t`foo`\n}\n", |
| 997 | }, |
| 998 | { |
| 999 | // { } expands to a proper block. |
| 1000 | name: "{ } empty block", |
| 1001 | input: "route { }", |
| 1002 | exactExpect: "route {\n}\n", |
| 1003 | }, |
| 1004 | { |
| 1005 | // Hash inside a double-quoted string is not a comment. |
| 1006 | name: "hash inside double-quoted string", |
| 1007 | input: `foo "bar#baz" quux`, |
| 1008 | exactExpect: "foo \"bar#baz\" quux\n", |
| 1009 | }, |
| 1010 | { |
| 1011 | // Hash inside a heredoc body is not a comment. |
| 1012 | name: "hash inside heredoc body", |
| 1013 | input: "x <<END\nfoo # not a comment\nEND\n", |
| 1014 | exactExpect: "x <<END\nfoo # not a comment\nEND\n", |
| 1015 | }, |
| 1016 | { |
| 1017 | // Escaped heredoc opener (\<<) is a regular token, not a real heredoc. |
| 1018 | name: "escaped heredoc \\<<", |
| 1019 | input: "block {\n\theredoc \\<<HEREDOC\n\trespond hello 200\n}\n", |
| 1020 | exactExpect: "block {\n\theredoc \\<<HEREDOC\n\trespond hello 200\n}\n", |
| 1021 | }, |
| 1022 | { |
| 1023 | // Heredoc whose marker appears as a substring of a body line (fooEND ≠ END). |
| 1024 | name: "heredoc marker as substring of body line", |
| 1025 | input: "x <<END\nfooEND\nEND\n", |
| 1026 | exactExpect: "x <<END\nfooEND\nEND\n", |
| 1027 | }, |
| 1028 | |
| 1029 | // ---- pathological: idempotency + no-panic only ---- |
| 1030 | |
| 1031 | { |
| 1032 | // Unbalanced: one more opening brace than closing braces. |
| 1033 | name: "unbalanced braces (too many open)", |
| 1034 | input: "a {\n\tb {\n\t\tc\n\t}\n", |
| 1035 | }, |
| 1036 | { |
| 1037 | // Unbalanced: stray closing brace. |
| 1038 | name: "unbalanced braces (too many close)", |
| 1039 | input: "a }\n", |
| 1040 | }, |
| 1041 | { |
| 1042 | // CRLF line endings must not break idempotency. |
| 1043 | name: "CRLF line endings", |
| 1044 | input: "site {\r\n\tfoo\r\n}\r\n", |
| 1045 | }, |
| 1046 | { |
| 1047 | // Unterminated double-quoted string. |
| 1048 | name: "unterminated double-quote", |
| 1049 | input: "foo \"unterminated", |
| 1050 | }, |
| 1051 | { |
| 1052 | // Unterminated backtick string. |
| 1053 | name: "unterminated backtick", |
| 1054 | input: "foo `unterminated", |
| 1055 | }, |
| 1056 | { |
| 1057 | // An empty quoted token followed by a lone opening quote. The first |
| 1058 | // pass falls back and appends the mandatory newline; on the second |
| 1059 | // pass that newline is swallowed by the still-unterminated quote, so |
| 1060 | // the unterminated token must be detected from the token itself, not |
| 1061 | // from whether appending a newline changes the stream. |
| 1062 | name: "empty quoted token followed by lone quote", |
| 1063 | input: "\"\"\"", |
| 1064 | }, |
| 1065 | { |
| 1066 | // Same shape with a backtick opening the unterminated token. |
| 1067 | name: "empty quoted token followed by lone backtick", |
| 1068 | input: "\"\"`", |
| 1069 | }, |
| 1070 | { |
| 1071 | // An escaped quote running to end-of-input after a closed backtick |
| 1072 | // token. Like the cases above, the second pass must still recognize |
| 1073 | // the swallowing token once the mandatory newline sits inside it. |
| 1074 | name: "empty backtick token followed by escaped quote", |
| 1075 | input: "``\\\"", |
| 1076 | }, |
| 1077 | { |
| 1078 | // A lone CR does not terminate a token, so it stays in the token's |
| 1079 | // verbatim source without meaning the token swallowed end-of-input. |
| 1080 | // Treating it as swallowed would suppress normal rendering here and |
| 1081 | // leave the double space uncollapsed on the first pass only. |
| 1082 | name: "double space with trailing lone CR", |
| 1083 | input: "0 0\r", |
| 1084 | exactExpect: "0 0\n", |
| 1085 | }, |
| 1086 | { |
| 1087 | // Trailing backslash (dangling escape — not a line continuation). |
| 1088 | name: "trailing backslash", |
| 1089 | input: "a b\\", |
| 1090 | }, |
| 1091 | { |
| 1092 | // UTF-8 BOM at the start of the file; the BOM should not survive into |
| 1093 | // the formatted output (it is part of the first token's raw source but |
| 1094 | // the formatter strips leading/trailing whitespace). |
| 1095 | name: "UTF-8 BOM prefix", |
| 1096 | input: "\xef\xbb\xbfsite {\n\tfoo\n}\n", |
| 1097 | }, |
| 1098 | { |
| 1099 | // NUL byte embedded in a token. |
| 1100 | name: "NUL byte in token", |
| 1101 | input: "site {\nfoo\x00bar\n}\n", |
| 1102 | }, |
| 1103 | { |
| 1104 | // Arbitrary control bytes in a token. |
| 1105 | name: "control bytes in token", |
| 1106 | input: "a\x01b\x02c", |
| 1107 | }, |
| 1108 | { |
| 1109 | // Lone carriage-return (CR without LF). The lexer treats \r as part of |
| 1110 | // the token raw text but strips it from the token's text field; the raw |
| 1111 | // bytes are emitted verbatim by the formatter. |
| 1112 | name: "lone CR (respond hello\\rworld)", |
| 1113 | input: "respond hello\rworld", |
| 1114 | }, |
| 1115 | } |
| 1116 | |
| 1117 | for _, c := range cases { |
| 1118 | t.Run(c.name, func(t *testing.T) { |
| 1119 | // No-panic: the call itself must complete. |
| 1120 | once := Format([]byte(c.input)) |
| 1121 | |
| 1122 | // Idempotency: a second pass must reproduce the first pass exactly. |
| 1123 | twice := Format(once) |
| 1124 | if !bytes.Equal(once, twice) { |
| 1125 | t.Errorf("not idempotent:\n once=%q\ntwice=%q", once, twice) |
| 1126 | } |
| 1127 | |
| 1128 | // Exact-output assertion for clearly-idiomatic cases. |
| 1129 | if c.exactExpect != "" && string(once) != c.exactExpect { |
| 1130 | t.Errorf("unexpected output:\n got %q\n want %q", string(once), c.exactExpect) |
| 1131 | } |
| 1132 | }) |
| 1133 | } |
| 1134 | } |
| 1135 | |
| 1136 | // TestFormattingDifferenceStableOnFormatted verifies that an already-formatted |
| 1137 | // config produces no diff from FormattingDifference (i.e. the formatter and the |
| 1138 | // diff checker agree that a well-formed file needs no changes). |
| 1139 | func TestFormattingDifferenceStableOnFormatted(t *testing.T) { |
| 1140 | in := []byte("site {\n\troot * /srv\n\tfile_server\n}\n") |
| 1141 | formatted := Format(in) |
| 1142 | if _, diff := FormattingDifference("Caddyfile", formatted); diff { |
| 1143 | t.Error("FormattingDifference reported a diff on already-formatted input") |
| 1144 | } |
| 1145 | } |
| 1146 | |
| 1147 | // sameStructure compares two parses by their per-segment token Text sequences. |
| 1148 | func sameStructure(a, b []ServerBlock) bool { |
| 1149 | seq := func(blocks []ServerBlock) []string { |
| 1150 | var s []string |
| 1151 | for _, blk := range blocks { |
| 1152 | s = append(s, "K") |
| 1153 | for _, k := range blk.Keys { |
| 1154 | s = append(s, k.Text) |
| 1155 | } |
| 1156 | for _, seg := range blk.Segments { |
| 1157 | s = append(s, "S") |
| 1158 | for _, tk := range seg { |
| 1159 | s = append(s, tk.Text) |
| 1160 | } |
| 1161 | } |
| 1162 | } |
| 1163 | return s |
| 1164 | } |
| 1165 | return reflect.DeepEqual(seq(a), seq(b)) |
| 1166 | } |