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