1FileCheck - Flexible pattern matching file verifier 2=================================================== 3 4SYNOPSIS 5-------- 6 7:program:`FileCheck` *match-filename* [*--check-prefix=XXX*] [*--strict-whitespace*] 8 9DESCRIPTION 10----------- 11 12:program:`FileCheck` reads two files (one from standard input, and one 13specified on the command line) and uses one to verify the other. This 14behavior is particularly useful for the testsuite, which wants to verify that 15the output of some tool (e.g. :program:`llc`) contains the expected information 16(for example, a movsd from esp or whatever is interesting). This is similar to 17using :program:`grep`, but it is optimized for matching multiple different 18inputs in one file in a specific order. 19 20The ``match-filename`` file specifies the file that contains the patterns to 21match. The file to verify is read from standard input unless the 22:option:`--input-file` option is used. 23 24OPTIONS 25------- 26 27.. option:: -help 28 29 Print a summary of command line options. 30 31.. option:: --check-prefix prefix 32 33 FileCheck searches the contents of ``match-filename`` for patterns to 34 match. By default, these patterns are prefixed with "``CHECK:``". 35 If you'd like to use a different prefix (e.g. because the same input 36 file is checking multiple different tool or options), the 37 :option:`--check-prefix` argument allows you to specify one or more 38 prefixes to match. Multiple prefixes are useful for tests which might 39 change for different run options, but most lines remain the same. 40 41.. option:: --check-prefixes prefix1,prefix2,... 42 43 An alias of :option:`--check-prefix` that allows multiple prefixes to be 44 specified as a comma separated list. 45 46.. option:: --input-file filename 47 48 File to check (defaults to stdin). 49 50.. option:: --match-full-lines 51 52 By default, FileCheck allows matches of anywhere on a line. This 53 option will require all positive matches to cover an entire 54 line. Leading and trailing whitespace is ignored, unless 55 :option:`--strict-whitespace` is also specified. (Note: negative 56 matches from ``CHECK-NOT`` are not affected by this option!) 57 58 Passing this option is equivalent to inserting ``{{^ *}}`` or 59 ``{{^}}`` before, and ``{{ *$}}`` or ``{{$}}`` after every positive 60 check pattern. 61 62.. option:: --strict-whitespace 63 64 By default, FileCheck canonicalizes input horizontal whitespace (spaces and 65 tabs) which causes it to ignore these differences (a space will match a tab). 66 The :option:`--strict-whitespace` argument disables this behavior. End-of-line 67 sequences are canonicalized to UNIX-style ``\n`` in all modes. 68 69.. option:: --implicit-check-not check-pattern 70 71 Adds implicit negative checks for the specified patterns between positive 72 checks. The option allows writing stricter tests without stuffing them with 73 ``CHECK-NOT``\ s. 74 75 For example, "``--implicit-check-not warning:``" can be useful when testing 76 diagnostic messages from tools that don't have an option similar to ``clang 77 -verify``. With this option FileCheck will verify that input does not contain 78 warnings not covered by any ``CHECK:`` patterns. 79 80.. option:: --enable-var-scope 81 82 Enables scope for regex variables. 83 84 Variables with names that start with ``$`` are considered global and 85 remain set throughout the file. 86 87 All other variables get undefined after each encountered ``CHECK-LABEL``. 88 89.. option:: -D<VAR=VALUE> 90 91 Sets a filecheck variable ``VAR`` with value ``VALUE`` that can be used in 92 ``CHECK:`` lines. 93 94.. option:: -version 95 96 Show the version number of this program. 97 98EXIT STATUS 99----------- 100 101If :program:`FileCheck` verifies that the file matches the expected contents, 102it exits with 0. Otherwise, if not, or if an error occurs, it will exit with a 103non-zero value. 104 105TUTORIAL 106-------- 107 108FileCheck is typically used from LLVM regression tests, being invoked on the RUN 109line of the test. A simple example of using FileCheck from a RUN line looks 110like this: 111 112.. code-block:: llvm 113 114 ; RUN: llvm-as < %s | llc -march=x86-64 | FileCheck %s 115 116This syntax says to pipe the current file ("``%s``") into ``llvm-as``, pipe 117that into ``llc``, then pipe the output of ``llc`` into ``FileCheck``. This 118means that FileCheck will be verifying its standard input (the llc output) 119against the filename argument specified (the original ``.ll`` file specified by 120"``%s``"). To see how this works, let's look at the rest of the ``.ll`` file 121(after the RUN line): 122 123.. code-block:: llvm 124 125 define void @sub1(i32* %p, i32 %v) { 126 entry: 127 ; CHECK: sub1: 128 ; CHECK: subl 129 %0 = tail call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %p, i32 %v) 130 ret void 131 } 132 133 define void @inc4(i64* %p) { 134 entry: 135 ; CHECK: inc4: 136 ; CHECK: incq 137 %0 = tail call i64 @llvm.atomic.load.add.i64.p0i64(i64* %p, i64 1) 138 ret void 139 } 140 141Here you can see some "``CHECK:``" lines specified in comments. Now you can 142see how the file is piped into ``llvm-as``, then ``llc``, and the machine code 143output is what we are verifying. FileCheck checks the machine code output to 144verify that it matches what the "``CHECK:``" lines specify. 145 146The syntax of the "``CHECK:``" lines is very simple: they are fixed strings that 147must occur in order. FileCheck defaults to ignoring horizontal whitespace 148differences (e.g. a space is allowed to match a tab) but otherwise, the contents 149of the "``CHECK:``" line is required to match some thing in the test file exactly. 150 151One nice thing about FileCheck (compared to grep) is that it allows merging 152test cases together into logical groups. For example, because the test above 153is checking for the "``sub1:``" and "``inc4:``" labels, it will not match 154unless there is a "``subl``" in between those labels. If it existed somewhere 155else in the file, that would not count: "``grep subl``" matches if "``subl``" 156exists anywhere in the file. 157 158The FileCheck -check-prefix option 159~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 160 161The FileCheck `-check-prefix` option allows multiple test 162configurations to be driven from one `.ll` file. This is useful in many 163circumstances, for example, testing different architectural variants with 164:program:`llc`. Here's a simple example: 165 166.. code-block:: llvm 167 168 ; RUN: llvm-as < %s | llc -mtriple=i686-apple-darwin9 -mattr=sse41 \ 169 ; RUN: | FileCheck %s -check-prefix=X32 170 ; RUN: llvm-as < %s | llc -mtriple=x86_64-apple-darwin9 -mattr=sse41 \ 171 ; RUN: | FileCheck %s -check-prefix=X64 172 173 define <4 x i32> @pinsrd_1(i32 %s, <4 x i32> %tmp) nounwind { 174 %tmp1 = insertelement <4 x i32>; %tmp, i32 %s, i32 1 175 ret <4 x i32> %tmp1 176 ; X32: pinsrd_1: 177 ; X32: pinsrd $1, 4(%esp), %xmm0 178 179 ; X64: pinsrd_1: 180 ; X64: pinsrd $1, %edi, %xmm0 181 } 182 183In this case, we're testing that we get the expected code generation with 184both 32-bit and 64-bit code generation. 185 186The "CHECK-NEXT:" directive 187~~~~~~~~~~~~~~~~~~~~~~~~~~~ 188 189Sometimes you want to match lines and would like to verify that matches 190happen on exactly consecutive lines with no other lines in between them. In 191this case, you can use "``CHECK:``" and "``CHECK-NEXT:``" directives to specify 192this. If you specified a custom check prefix, just use "``<PREFIX>-NEXT:``". 193For example, something like this works as you'd expect: 194 195.. code-block:: llvm 196 197 define void @t2(<2 x double>* %r, <2 x double>* %A, double %B) { 198 %tmp3 = load <2 x double>* %A, align 16 199 %tmp7 = insertelement <2 x double> undef, double %B, i32 0 200 %tmp9 = shufflevector <2 x double> %tmp3, 201 <2 x double> %tmp7, 202 <2 x i32> < i32 0, i32 2 > 203 store <2 x double> %tmp9, <2 x double>* %r, align 16 204 ret void 205 206 ; CHECK: t2: 207 ; CHECK: movl 8(%esp), %eax 208 ; CHECK-NEXT: movapd (%eax), %xmm0 209 ; CHECK-NEXT: movhpd 12(%esp), %xmm0 210 ; CHECK-NEXT: movl 4(%esp), %eax 211 ; CHECK-NEXT: movapd %xmm0, (%eax) 212 ; CHECK-NEXT: ret 213 } 214 215"``CHECK-NEXT:``" directives reject the input unless there is exactly one 216newline between it and the previous directive. A "``CHECK-NEXT:``" cannot be 217the first directive in a file. 218 219The "CHECK-SAME:" directive 220~~~~~~~~~~~~~~~~~~~~~~~~~~~ 221 222Sometimes you want to match lines and would like to verify that matches happen 223on the same line as the previous match. In this case, you can use "``CHECK:``" 224and "``CHECK-SAME:``" directives to specify this. If you specified a custom 225check prefix, just use "``<PREFIX>-SAME:``". 226 227"``CHECK-SAME:``" is particularly powerful in conjunction with "``CHECK-NOT:``" 228(described below). 229 230For example, the following works like you'd expect: 231 232.. code-block:: llvm 233 234 !0 = !DILocation(line: 5, scope: !1, inlinedAt: !2) 235 236 ; CHECK: !DILocation(line: 5, 237 ; CHECK-NOT: column: 238 ; CHECK-SAME: scope: ![[SCOPE:[0-9]+]] 239 240"``CHECK-SAME:``" directives reject the input if there are any newlines between 241it and the previous directive. A "``CHECK-SAME:``" cannot be the first 242directive in a file. 243 244The "CHECK-NOT:" directive 245~~~~~~~~~~~~~~~~~~~~~~~~~~ 246 247The "``CHECK-NOT:``" directive is used to verify that a string doesn't occur 248between two matches (or before the first match, or after the last match). For 249example, to verify that a load is removed by a transformation, a test like this 250can be used: 251 252.. code-block:: llvm 253 254 define i8 @coerce_offset0(i32 %V, i32* %P) { 255 store i32 %V, i32* %P 256 257 %P2 = bitcast i32* %P to i8* 258 %P3 = getelementptr i8* %P2, i32 2 259 260 %A = load i8* %P3 261 ret i8 %A 262 ; CHECK: @coerce_offset0 263 ; CHECK-NOT: load 264 ; CHECK: ret i8 265 } 266 267The "CHECK-DAG:" directive 268~~~~~~~~~~~~~~~~~~~~~~~~~~ 269 270If it's necessary to match strings that don't occur in a strictly sequential 271order, "``CHECK-DAG:``" could be used to verify them between two matches (or 272before the first match, or after the last match). For example, clang emits 273vtable globals in reverse order. Using ``CHECK-DAG:``, we can keep the checks 274in the natural order: 275 276.. code-block:: c++ 277 278 // RUN: %clang_cc1 %s -emit-llvm -o - | FileCheck %s 279 280 struct Foo { virtual void method(); }; 281 Foo f; // emit vtable 282 // CHECK-DAG: @_ZTV3Foo = 283 284 struct Bar { virtual void method(); }; 285 Bar b; 286 // CHECK-DAG: @_ZTV3Bar = 287 288``CHECK-NOT:`` directives could be mixed with ``CHECK-DAG:`` directives to 289exclude strings between the surrounding ``CHECK-DAG:`` directives. As a result, 290the surrounding ``CHECK-DAG:`` directives cannot be reordered, i.e. all 291occurrences matching ``CHECK-DAG:`` before ``CHECK-NOT:`` must not fall behind 292occurrences matching ``CHECK-DAG:`` after ``CHECK-NOT:``. For example, 293 294.. code-block:: llvm 295 296 ; CHECK-DAG: BEFORE 297 ; CHECK-NOT: NOT 298 ; CHECK-DAG: AFTER 299 300This case will reject input strings where ``BEFORE`` occurs after ``AFTER``. 301 302With captured variables, ``CHECK-DAG:`` is able to match valid topological 303orderings of a DAG with edges from the definition of a variable to its use. 304It's useful, e.g., when your test cases need to match different output 305sequences from the instruction scheduler. For example, 306 307.. code-block:: llvm 308 309 ; CHECK-DAG: add [[REG1:r[0-9]+]], r1, r2 310 ; CHECK-DAG: add [[REG2:r[0-9]+]], r3, r4 311 ; CHECK: mul r5, [[REG1]], [[REG2]] 312 313In this case, any order of that two ``add`` instructions will be allowed. 314 315If you are defining `and` using variables in the same ``CHECK-DAG:`` block, 316be aware that the definition rule can match `after` its use. 317 318So, for instance, the code below will pass: 319 320.. code-block:: text 321 322 ; CHECK-DAG: vmov.32 [[REG2:d[0-9]+]][0] 323 ; CHECK-DAG: vmov.32 [[REG2]][1] 324 vmov.32 d0[1] 325 vmov.32 d0[0] 326 327While this other code, will not: 328 329.. code-block:: text 330 331 ; CHECK-DAG: vmov.32 [[REG2:d[0-9]+]][0] 332 ; CHECK-DAG: vmov.32 [[REG2]][1] 333 vmov.32 d1[1] 334 vmov.32 d0[0] 335 336While this can be very useful, it's also dangerous, because in the case of 337register sequence, you must have a strong order (read before write, copy before 338use, etc). If the definition your test is looking for doesn't match (because 339of a bug in the compiler), it may match further away from the use, and mask 340real bugs away. 341 342In those cases, to enforce the order, use a non-DAG directive between DAG-blocks. 343 344The "CHECK-LABEL:" directive 345~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 346 347Sometimes in a file containing multiple tests divided into logical blocks, one 348or more ``CHECK:`` directives may inadvertently succeed by matching lines in a 349later block. While an error will usually eventually be generated, the check 350flagged as causing the error may not actually bear any relationship to the 351actual source of the problem. 352 353In order to produce better error messages in these cases, the "``CHECK-LABEL:``" 354directive can be used. It is treated identically to a normal ``CHECK`` 355directive except that FileCheck makes an additional assumption that a line 356matched by the directive cannot also be matched by any other check present in 357``match-filename``; this is intended to be used for lines containing labels or 358other unique identifiers. Conceptually, the presence of ``CHECK-LABEL`` divides 359the input stream into separate blocks, each of which is processed independently, 360preventing a ``CHECK:`` directive in one block matching a line in another block. 361If ``--enable-var-scope`` is in effect, all local variables are cleared at the 362beginning of the block. 363 364For example, 365 366.. code-block:: llvm 367 368 define %struct.C* @C_ctor_base(%struct.C* %this, i32 %x) { 369 entry: 370 ; CHECK-LABEL: C_ctor_base: 371 ; CHECK: mov [[SAVETHIS:r[0-9]+]], r0 372 ; CHECK: bl A_ctor_base 373 ; CHECK: mov r0, [[SAVETHIS]] 374 %0 = bitcast %struct.C* %this to %struct.A* 375 %call = tail call %struct.A* @A_ctor_base(%struct.A* %0) 376 %1 = bitcast %struct.C* %this to %struct.B* 377 %call2 = tail call %struct.B* @B_ctor_base(%struct.B* %1, i32 %x) 378 ret %struct.C* %this 379 } 380 381 define %struct.D* @D_ctor_base(%struct.D* %this, i32 %x) { 382 entry: 383 ; CHECK-LABEL: D_ctor_base: 384 385The use of ``CHECK-LABEL:`` directives in this case ensures that the three 386``CHECK:`` directives only accept lines corresponding to the body of the 387``@C_ctor_base`` function, even if the patterns match lines found later in 388the file. Furthermore, if one of these three ``CHECK:`` directives fail, 389FileCheck will recover by continuing to the next block, allowing multiple test 390failures to be detected in a single invocation. 391 392There is no requirement that ``CHECK-LABEL:`` directives contain strings that 393correspond to actual syntactic labels in a source or output language: they must 394simply uniquely match a single line in the file being verified. 395 396``CHECK-LABEL:`` directives cannot contain variable definitions or uses. 397 398FileCheck Pattern Matching Syntax 399~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 400 401All FileCheck directives take a pattern to match. 402For most uses of FileCheck, fixed string matching is perfectly sufficient. For 403some things, a more flexible form of matching is desired. To support this, 404FileCheck allows you to specify regular expressions in matching strings, 405surrounded by double braces: ``{{yourregex}}``. FileCheck implements a POSIX 406regular expression matcher; it supports Extended POSIX regular expressions 407(ERE). Because we want to use fixed string matching for a majority of what we 408do, FileCheck has been designed to support mixing and matching fixed string 409matching with regular expressions. This allows you to write things like this: 410 411.. code-block:: llvm 412 413 ; CHECK: movhpd {{[0-9]+}}(%esp), {{%xmm[0-7]}} 414 415In this case, any offset from the ESP register will be allowed, and any xmm 416register will be allowed. 417 418Because regular expressions are enclosed with double braces, they are 419visually distinct, and you don't need to use escape characters within the double 420braces like you would in C. In the rare case that you want to match double 421braces explicitly from the input, you can use something ugly like 422``{{[{][{]}}`` as your pattern. 423 424FileCheck Variables 425~~~~~~~~~~~~~~~~~~~ 426 427It is often useful to match a pattern and then verify that it occurs again 428later in the file. For codegen tests, this can be useful to allow any register, 429but verify that that register is used consistently later. To do this, 430:program:`FileCheck` allows named variables to be defined and substituted into 431patterns. Here is a simple example: 432 433.. code-block:: llvm 434 435 ; CHECK: test5: 436 ; CHECK: notw [[REGISTER:%[a-z]+]] 437 ; CHECK: andw {{.*}}[[REGISTER]] 438 439The first check line matches a regex ``%[a-z]+`` and captures it into the 440variable ``REGISTER``. The second line verifies that whatever is in 441``REGISTER`` occurs later in the file after an "``andw``". :program:`FileCheck` 442variable references are always contained in ``[[ ]]`` pairs, and their names can 443be formed with the regex ``[a-zA-Z_][a-zA-Z0-9_]*``. If a colon follows the name, 444then it is a definition of the variable; otherwise, it is a use. 445 446:program:`FileCheck` variables can be defined multiple times, and uses always 447get the latest value. Variables can also be used later on the same line they 448were defined on. For example: 449 450.. code-block:: llvm 451 452 ; CHECK: op [[REG:r[0-9]+]], [[REG]] 453 454Can be useful if you want the operands of ``op`` to be the same register, 455and don't care exactly which register it is. 456 457If ``--enable-var-scope`` is in effect, variables with names that 458start with ``$`` are considered to be global. All others variables are 459local. All local variables get undefined at the beginning of each 460CHECK-LABEL block. Global variables are not affected by CHECK-LABEL. 461This makes it easier to ensure that individual tests are not affected 462by variables set in preceding tests. 463 464FileCheck Expressions 465~~~~~~~~~~~~~~~~~~~~~ 466 467Sometimes there's a need to verify output which refers line numbers of the 468match file, e.g. when testing compiler diagnostics. This introduces a certain 469fragility of the match file structure, as "``CHECK:``" lines contain absolute 470line numbers in the same file, which have to be updated whenever line numbers 471change due to text addition or deletion. 472 473To support this case, FileCheck allows using ``[[@LINE]]``, 474``[[@LINE+<offset>]]``, ``[[@LINE-<offset>]]`` expressions in patterns. These 475expressions expand to a number of the line where a pattern is located (with an 476optional integer offset). 477 478This way match patterns can be put near the relevant test lines and include 479relative line number references, for example: 480 481.. code-block:: c++ 482 483 // CHECK: test.cpp:[[@LINE+4]]:6: error: expected ';' after top level declarator 484 // CHECK-NEXT: {{^int a}} 485 // CHECK-NEXT: {{^ \^}} 486 // CHECK-NEXT: {{^ ;}} 487 int a 488 489Matching Newline Characters 490~~~~~~~~~~~~~~~~~~~~~~~~~~~ 491 492To match newline characters in regular expressions the character class 493``[[:space:]]`` can be used. For example, the following pattern: 494 495.. code-block:: c++ 496 497 // CHECK: DW_AT_location [DW_FORM_sec_offset] ([[DLOC:0x[0-9a-f]+]]){{[[:space:]].*}}"intd" 498 499matches output of the form (from llvm-dwarfdump): 500 501.. code-block:: text 502 503 DW_AT_location [DW_FORM_sec_offset] (0x00000233) 504 DW_AT_name [DW_FORM_strp] ( .debug_str[0x000000c9] = "intd") 505 506letting us set the :program:`FileCheck` variable ``DLOC`` to the desired value 507``0x00000233``, extracted from the line immediately preceding "``intd``". 508