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