1FILECHECK(1) LLVM FILECHECK(1)
2
3
4
6 FileCheck - Flexible pattern matching file verifier
7
9 FileCheck match-filename [--check-prefix=XXX] [--strict-whitespace]
10
12 FileCheck reads two files (one from standard input, and one specified
13 on the command line) and uses one to verify the other. This behavior
14 is particularly useful for the testsuite, which wants to verify that
15 the output of some tool (e.g. llc) contains the expected information
16 (for example, a movsd from esp or whatever is interesting). This is
17 similar to using grep, but it is optimized for matching multiple dif‐
18 ferent inputs in one file in a specific order.
19
20 The match-filename file specifies the file that contains the patterns
21 to match. The file to verify is read from standard input unless the
22 --input-file option is used.
23
25 Options are parsed from the environment variable FILECHECK_OPTS and
26 from the command line.
27
28 -help Print a summary of command line options.
29
30 --check-prefix prefix
31 FileCheck searches the contents of match-filename for patterns
32 to match. By default, these patterns are prefixed with
33 "CHECK:". If you'd like to use a different prefix (e.g. because
34 the same input file is checking multiple different tool or
35 options), the --check-prefix argument allows you to specify one
36 or more prefixes to match. Multiple prefixes are useful for
37 tests which might change for different run options, but most
38 lines remain the same.
39
40 --check-prefixes prefix1,prefix2,...
41 An alias of --check-prefix that allows multiple prefixes to be
42 specified as a comma separated list.
43
44 --input-file filename
45 File to check (defaults to stdin).
46
47 --match-full-lines
48 By default, FileCheck allows matches of anywhere on a line. This
49 option will require all positive matches to cover an entire
50 line. Leading and trailing whitespace is ignored, unless
51 --strict-whitespace is also specified. (Note: negative matches
52 from CHECK-NOT are not affected by this option!)
53
54 Passing this option is equivalent to inserting {{^ *}} or {{^}}
55 before, and {{ *$}} or {{$}} after every positive check pattern.
56
57 --strict-whitespace
58 By default, FileCheck canonicalizes input horizontal whitespace
59 (spaces and tabs) which causes it to ignore these differences (a
60 space will match a tab). The --strict-whitespace argument dis‐
61 ables this behavior. End-of-line sequences are canonicalized to
62 UNIX-style \n in all modes.
63
64 --implicit-check-not check-pattern
65 Adds implicit negative checks for the specified patterns between
66 positive checks. The option allows writing stricter tests with‐
67 out stuffing them with CHECK-NOTs.
68
69 For example, "--implicit-check-not warning:" can be useful when
70 testing diagnostic messages from tools that don't have an option
71 similar to clang -verify. With this option FileCheck will verify
72 that input does not contain warnings not covered by any CHECK:
73 patterns.
74
75 --dump-input <mode>
76 Dump input to stderr, adding annotations representing currently
77 enabled diagnostics. Do this either 'always', on 'fail', or
78 'never'. Specify 'help' to explain the dump format and quit.
79
80 --dump-input-on-failure
81 When the check fails, dump all of the original input. This
82 option is deprecated in favor of --dump-input=fail.
83
84 --enable-var-scope
85 Enables scope for regex variables.
86
87 Variables with names that start with $ are considered global and
88 remain set throughout the file.
89
90 All other variables get undefined after each encountered
91 CHECK-LABEL.
92
93 -D<VAR=VALUE>
94 Sets a filecheck variable VAR with value VALUE that can be used
95 in CHECK: lines.
96
97 -version
98 Show the version number of this program.
99
100 -v Print directive pattern matches.
101
102 -vv Print information helpful in diagnosing internal FileCheck
103 issues, such as discarded overlapping CHECK-DAG: matches,
104 implicit EOF pattern matches, and CHECK-NOT: patterns that do
105 not have matches. Implies -v.
106
107 --allow-deprecated-dag-overlap
108 Enable overlapping among matches in a group of consecutive
109 CHECK-DAG: directives. This option is deprecated and is only
110 provided for convenience as old tests are migrated to the new
111 non-overlapping CHECK-DAG: implementation.
112
113 --color
114 Use colors in output (autodetected by default).
115
117 If FileCheck verifies that the file matches the expected contents, it
118 exits with 0. Otherwise, if not, or if an error occurs, it will exit
119 with a non-zero value.
120
122 FileCheck is typically used from LLVM regression tests, being invoked
123 on the RUN line of the test. A simple example of using FileCheck from
124 a RUN line looks like this:
125
126 ; RUN: llvm-as < %s | llc -march=x86-64 | FileCheck %s
127
128 This syntax says to pipe the current file ("%s") into llvm-as, pipe
129 that into llc, then pipe the output of llc into FileCheck. This means
130 that FileCheck will be verifying its standard input (the llc output)
131 against the filename argument specified (the original .ll file speci‐
132 fied by "%s"). To see how this works, let's look at the rest of the
133 .ll file (after the RUN line):
134
135 define void @sub1(i32* %p, i32 %v) {
136 entry:
137 ; CHECK: sub1:
138 ; CHECK: subl
139 %0 = tail call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %p, i32 %v)
140 ret void
141 }
142
143 define void @inc4(i64* %p) {
144 entry:
145 ; CHECK: inc4:
146 ; CHECK: incq
147 %0 = tail call i64 @llvm.atomic.load.add.i64.p0i64(i64* %p, i64 1)
148 ret void
149 }
150
151 Here you can see some "CHECK:" lines specified in comments. Now you
152 can see how the file is piped into llvm-as, then llc, and the machine
153 code output is what we are verifying. FileCheck checks the machine
154 code output to verify that it matches what the "CHECK:" lines specify.
155
156 The syntax of the "CHECK:" lines is very simple: they are fixed strings
157 that must occur in order. FileCheck defaults to ignoring horizontal
158 whitespace differences (e.g. a space is allowed to match a tab) but
159 otherwise, the contents of the "CHECK:" line is required to match some
160 thing in the test file exactly.
161
162 One nice thing about FileCheck (compared to grep) is that it allows
163 merging test cases together into logical groups. For example, because
164 the test above is checking for the "sub1:" and "inc4:" labels, it will
165 not match unless there is a "subl" in between those labels. If it
166 existed somewhere else in the file, that would not count: "grep subl"
167 matches if "subl" exists anywhere in the file.
168
169 The FileCheck -check-prefix option
170 The FileCheck -check-prefix option allows multiple test configurations
171 to be driven from one .ll file. This is useful in many circumstances,
172 for example, testing different architectural variants with llc. Here's
173 a simple example:
174
175 ; RUN: llvm-as < %s | llc -mtriple=i686-apple-darwin9 -mattr=sse41 \
176 ; RUN: | FileCheck %s -check-prefix=X32
177 ; RUN: llvm-as < %s | llc -mtriple=x86_64-apple-darwin9 -mattr=sse41 \
178 ; RUN: | FileCheck %s -check-prefix=X64
179
180 define <4 x i32> @pinsrd_1(i32 %s, <4 x i32> %tmp) nounwind {
181 %tmp1 = insertelement <4 x i32>; %tmp, i32 %s, i32 1
182 ret <4 x i32> %tmp1
183 ; X32: pinsrd_1:
184 ; X32: pinsrd $1, 4(%esp), %xmm0
185
186 ; X64: pinsrd_1:
187 ; X64: pinsrd $1, %edi, %xmm0
188 }
189
190 In this case, we're testing that we get the expected code generation
191 with both 32-bit and 64-bit code generation.
192
193 The CHECK-NEXT: directive
194 Sometimes you want to match lines and would like to verify that matches
195 happen on exactly consecutive lines with no other lines in between
196 them. In this case, you can use "CHECK:" and "CHECK-NEXT:" directives
197 to specify this. If you specified a custom check prefix, just use
198 "<PREFIX>-NEXT:". For example, something like this works as you'd
199 expect:
200
201 define void @t2(<2 x double>* %r, <2 x double>* %A, double %B) {
202 %tmp3 = load <2 x double>* %A, align 16
203 %tmp7 = insertelement <2 x double> undef, double %B, i32 0
204 %tmp9 = shufflevector <2 x double> %tmp3,
205 <2 x double> %tmp7,
206 <2 x i32> < i32 0, i32 2 >
207 store <2 x double> %tmp9, <2 x double>* %r, align 16
208 ret void
209
210 ; CHECK: t2:
211 ; CHECK: movl 8(%esp), %eax
212 ; CHECK-NEXT: movapd (%eax), %xmm0
213 ; CHECK-NEXT: movhpd 12(%esp), %xmm0
214 ; CHECK-NEXT: movl 4(%esp), %eax
215 ; CHECK-NEXT: movapd %xmm0, (%eax)
216 ; CHECK-NEXT: ret
217 }
218
219 "CHECK-NEXT:" directives reject the input unless there is exactly one
220 newline between it and the previous directive. A "CHECK-NEXT:" cannot
221 be the first directive in a file.
222
223 The CHECK-SAME: directive
224 Sometimes you want to match lines and would like to verify that matches
225 happen on the same line as the previous match. In this case, you can
226 use "CHECK:" and "CHECK-SAME:" directives to specify this. If you
227 specified a custom check prefix, just use "<PREFIX>-SAME:".
228
229 "CHECK-SAME:" is particularly powerful in conjunction with "CHECK-NOT:"
230 (described below).
231
232 For example, the following works like you'd expect:
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
241 between it and the previous directive. A "CHECK-SAME:" cannot be the
242 first directive in a file.
243
244 The CHECK-EMPTY: directive
245 If you need to check that the next line has nothing on it, not even
246 whitespace, you can use the "CHECK-EMPTY:" directive.
247
248 declare void @foo()
249
250 declare void @bar()
251 ; CHECK: foo
252 ; CHECK-EMPTY:
253 ; CHECK-NEXT: bar
254
255 Just like "CHECK-NEXT:" the directive will fail if there is more than
256 one newline before it finds the next blank line, and it cannot be the
257 first directive in a file.
258
259 The CHECK-NOT: directive
260 The "CHECK-NOT:" directive is used to verify that a string doesn't
261 occur between two matches (or before the first match, or after the last
262 match). For example, to verify that a load is removed by a transforma‐
263 tion, a test like this can be used:
264
265 define i8 @coerce_offset0(i32 %V, i32* %P) {
266 store i32 %V, i32* %P
267
268 %P2 = bitcast i32* %P to i8*
269 %P3 = getelementptr i8* %P2, i32 2
270
271 %A = load i8* %P3
272 ret i8 %A
273 ; CHECK: @coerce_offset0
274 ; CHECK-NOT: load
275 ; CHECK: ret i8
276 }
277
278 The CHECK-COUNT: directive
279 If you need to match multiple lines with the same pattern over and over
280 again you can repeat a plain CHECK: as many times as needed. If that
281 looks too boring you can instead use a counted check
282 "CHECK-COUNT-<num>:", where <num> is a positive decimal number. It will
283 match the pattern exactly <num> times, no more and no less. If you
284 specified a custom check prefix, just use "<PREFIX>-COUNT-<num>:" for
285 the same effect. Here is a simple example:
286
287 Loop at depth 1
288 Loop at depth 1
289 Loop at depth 1
290 Loop at depth 1
291 Loop at depth 2
292 Loop at depth 3
293
294 ; CHECK-COUNT-6: Loop at depth {{[0-9]+}}
295 ; CHECK-NOT: Loop at depth {{[0-9]+}}
296
297 The CHECK-DAG: directive
298 If it's necessary to match strings that don't occur in a strictly
299 sequential order, "CHECK-DAG:" could be used to verify them between two
300 matches (or before the first match, or after the last match). For exam‐
301 ple, clang emits vtable globals in reverse order. Using CHECK-DAG:, we
302 can keep the checks in the natural order:
303
304 // RUN: %clang_cc1 %s -emit-llvm -o - | FileCheck %s
305
306 struct Foo { virtual void method(); };
307 Foo f; // emit vtable
308 // CHECK-DAG: @_ZTV3Foo =
309
310 struct Bar { virtual void method(); };
311 Bar b;
312 // CHECK-DAG: @_ZTV3Bar =
313
314 CHECK-NOT: directives could be mixed with CHECK-DAG: directives to
315 exclude strings between the surrounding CHECK-DAG: directives. As a
316 result, the surrounding CHECK-DAG: directives cannot be reordered, i.e.
317 all occurrences matching CHECK-DAG: before CHECK-NOT: must not fall
318 behind occurrences matching CHECK-DAG: after CHECK-NOT:. For example,
319
320 ; CHECK-DAG: BEFORE
321 ; CHECK-NOT: NOT
322 ; CHECK-DAG: AFTER
323
324 This case will reject input strings where BEFORE occurs after AFTER.
325
326 With captured variables, CHECK-DAG: is able to match valid topological
327 orderings of a DAG with edges from the definition of a variable to its
328 use. It's useful, e.g., when your test cases need to match different
329 output sequences from the instruction scheduler. For example,
330
331 ; CHECK-DAG: add [[REG1:r[0-9]+]], r1, r2
332 ; CHECK-DAG: add [[REG2:r[0-9]+]], r3, r4
333 ; CHECK: mul r5, [[REG1]], [[REG2]]
334
335 In this case, any order of that two add instructions will be allowed.
336
337 If you are defining and using variables in the same CHECK-DAG: block,
338 be aware that the definition rule can match after its use.
339
340 So, for instance, the code below will pass:
341
342 ; CHECK-DAG: vmov.32 [[REG2:d[0-9]+]][0]
343 ; CHECK-DAG: vmov.32 [[REG2]][1]
344 vmov.32 d0[1]
345 vmov.32 d0[0]
346
347 While this other code, will not:
348
349 ; CHECK-DAG: vmov.32 [[REG2:d[0-9]+]][0]
350 ; CHECK-DAG: vmov.32 [[REG2]][1]
351 vmov.32 d1[1]
352 vmov.32 d0[0]
353
354 While this can be very useful, it's also dangerous, because in the case
355 of register sequence, you must have a strong order (read before write,
356 copy before use, etc). If the definition your test is looking for
357 doesn't match (because of a bug in the compiler), it may match further
358 away from the use, and mask real bugs away.
359
360 In those cases, to enforce the order, use a non-DAG directive between
361 DAG-blocks.
362
363 A CHECK-DAG: directive skips matches that overlap the matches of any
364 preceding CHECK-DAG: directives in the same CHECK-DAG: block. Not only
365 is this non-overlapping behavior consistent with other directives, but
366 it's also necessary to handle sets of non-unique strings or patterns.
367 For example, the following directives look for unordered log entries
368 for two tasks in a parallel program, such as the OpenMP runtime:
369
370 // CHECK-DAG: [[THREAD_ID:[0-9]+]]: task_begin
371 // CHECK-DAG: [[THREAD_ID]]: task_end
372 //
373 // CHECK-DAG: [[THREAD_ID:[0-9]+]]: task_begin
374 // CHECK-DAG: [[THREAD_ID]]: task_end
375
376 The second pair of directives is guaranteed not to match the same log
377 entries as the first pair even though the patterns are identical and
378 even if the text of the log entries is identical because the thread ID
379 manages to be reused.
380
381 The CHECK-LABEL: directive
382 Sometimes in a file containing multiple tests divided into logical
383 blocks, one or more CHECK: directives may inadvertently succeed by
384 matching lines in a later block. While an error will usually eventually
385 be generated, the check flagged as causing the error may not actually
386 bear any relationship to the actual source of the problem.
387
388 In order to produce better error messages in these cases, the
389 "CHECK-LABEL:" directive can be used. It is treated identically to a
390 normal CHECK directive except that FileCheck makes an additional
391 assumption that a line matched by the directive cannot also be matched
392 by any other check present in match-filename; this is intended to be
393 used for lines containing labels or other unique identifiers. Conceptu‐
394 ally, the presence of CHECK-LABEL divides the input stream into sepa‐
395 rate blocks, each of which is processed independently, preventing a
396 CHECK: directive in one block matching a line in another block. If
397 --enable-var-scope is in effect, all local variables are cleared at the
398 beginning of the block.
399
400 For example,
401
402 define %struct.C* @C_ctor_base(%struct.C* %this, i32 %x) {
403 entry:
404 ; CHECK-LABEL: C_ctor_base:
405 ; CHECK: mov [[SAVETHIS:r[0-9]+]], r0
406 ; CHECK: bl A_ctor_base
407 ; CHECK: mov r0, [[SAVETHIS]]
408 %0 = bitcast %struct.C* %this to %struct.A*
409 %call = tail call %struct.A* @A_ctor_base(%struct.A* %0)
410 %1 = bitcast %struct.C* %this to %struct.B*
411 %call2 = tail call %struct.B* @B_ctor_base(%struct.B* %1, i32 %x)
412 ret %struct.C* %this
413 }
414
415 define %struct.D* @D_ctor_base(%struct.D* %this, i32 %x) {
416 entry:
417 ; CHECK-LABEL: D_ctor_base:
418
419 The use of CHECK-LABEL: directives in this case ensures that the three
420 CHECK: directives only accept lines corresponding to the body of the
421 @C_ctor_base function, even if the patterns match lines found later in
422 the file. Furthermore, if one of these three CHECK: directives fail,
423 FileCheck will recover by continuing to the next block, allowing multi‐
424 ple test failures to be detected in a single invocation.
425
426 There is no requirement that CHECK-LABEL: directives contain strings
427 that correspond to actual syntactic labels in a source or output lan‐
428 guage: they must simply uniquely match a single line in the file being
429 verified.
430
431 CHECK-LABEL: directives cannot contain variable definitions or uses.
432
433 FileCheck Pattern Matching Syntax
434 All FileCheck directives take a pattern to match. For most uses of
435 FileCheck, fixed string matching is perfectly sufficient. For some
436 things, a more flexible form of matching is desired. To support this,
437 FileCheck allows you to specify regular expressions in matching
438 strings, surrounded by double braces: {{yourregex}}. FileCheck imple‐
439 ments a POSIX regular expression matcher; it supports Extended POSIX
440 regular expressions (ERE). Because we want to use fixed string matching
441 for a majority of what we do, FileCheck has been designed to support
442 mixing and matching fixed string matching with regular expressions.
443 This allows you to write things like this:
444
445 ; CHECK: movhpd {{[0-9]+}}(%esp), {{%xmm[0-7]}}
446
447 In this case, any offset from the ESP register will be allowed, and any
448 xmm register will be allowed.
449
450 Because regular expressions are enclosed with double braces, they are
451 visually distinct, and you don't need to use escape characters within
452 the double braces like you would in C. In the rare case that you want
453 to match double braces explicitly from the input, you can use something
454 ugly like {{[{][{]}} as your pattern.
455
456 FileCheck Variables
457 It is often useful to match a pattern and then verify that it occurs
458 again later in the file. For codegen tests, this can be useful to
459 allow any register, but verify that that register is used consistently
460 later. To do this, FileCheck allows named variables to be defined and
461 substituted into patterns. Here is a simple example:
462
463 ; CHECK: test5:
464 ; CHECK: notw [[REGISTER:%[a-z]+]]
465 ; CHECK: andw {{.*}}[[REGISTER]]
466
467 The first check line matches a regex %[a-z]+ and captures it into the
468 variable REGISTER. The second line verifies that whatever is in REGIS‐
469 TER occurs later in the file after an "andw". FileCheck variable ref‐
470 erences are always contained in [[ ]] pairs, and their names can be
471 formed with the regex [a-zA-Z_][a-zA-Z0-9_]*. If a colon follows the
472 name, then it is a definition of the variable; otherwise, it is a use.
473
474 FileCheck variables can be defined multiple times, and uses always get
475 the latest value. Variables can also be used later on the same line
476 they were defined on. For example:
477
478 ; CHECK: op [[REG:r[0-9]+]], [[REG]]
479
480 Can be useful if you want the operands of op to be the same register,
481 and don't care exactly which register it is.
482
483 If --enable-var-scope is in effect, variables with names that start
484 with $ are considered to be global. All others variables are local.
485 All local variables get undefined at the beginning of each CHECK-LABEL
486 block. Global variables are not affected by CHECK-LABEL. This makes it
487 easier to ensure that individual tests are not affected by variables
488 set in preceding tests.
489
490 FileCheck Expressions
491 Sometimes there's a need to verify output which refers line numbers of
492 the match file, e.g. when testing compiler diagnostics. This intro‐
493 duces a certain fragility of the match file structure, as "CHECK:"
494 lines contain absolute line numbers in the same file, which have to be
495 updated whenever line numbers change due to text addition or deletion.
496
497 To support this case, FileCheck allows using [[@LINE]], [[@LINE+<off‐
498 set>]], [[@LINE-<offset>]] expressions in patterns. These expressions
499 expand to a number of the line where a pattern is located (with an
500 optional integer offset).
501
502 This way match patterns can be put near the relevant test lines and
503 include relative line number references, for example:
504
505 // CHECK: test.cpp:[[@LINE+4]]:6: error: expected ';' after top level declarator
506 // CHECK-NEXT: {{^int a}}
507 // CHECK-NEXT: {{^ \^}}
508 // CHECK-NEXT: {{^ ;}}
509 int a
510
511 Matching Newline Characters
512 To match newline characters in regular expressions the character class
513 [[:space:]] can be used. For example, the following pattern:
514
515 // CHECK: DW_AT_location [DW_FORM_sec_offset] ([[DLOC:0x[0-9a-f]+]]){{[[:space:]].*}}"intd"
516
517 matches output of the form (from llvm-dwarfdump):
518
519 DW_AT_location [DW_FORM_sec_offset] (0x00000233)
520 DW_AT_name [DW_FORM_strp] ( .debug_str[0x000000c9] = "intd")
521
522 letting us set the FileCheck variable DLOC to the desired value
523 0x00000233, extracted from the line immediately preceding "intd".
524
526 Maintained by the LLVM Team (https://llvm.org/).
527
529 2003-2019, LLVM Project
530
531
532
533
5348 2019-11-13 FILECHECK(1)