FILECHECK(1)

📦 version 23.1.3, in centos-stream 9

✨ NAME ✨

okay so this is FILECHECK(1) — FileCheck: flexible pattern matching file verifier, brought to you by LLVM. That's the name section, bestie, and honestly it already ate. 💅

🌈 SYNOPSIS 🌈

FileCheck match-filename [--check-prefixes=XXX] [--strict-whitespace]

That's the whole shape of it. Cute, right? 🎀

💅 DESCRIPTION 💅

FileCheck reads two files — one from standard input, and one specified on the command line — and uses one to verify the other. This behavior is particularly useful for the testsuite, which wants to verify that the output of some tool (e.g. llc) contains the expected information (for example, a movsd from esp or whatever is interesting). It's similar to using grep, but it's optimized for matching multiple different inputs in one file in a specific order. fr fr, that ordering is the whole point. 💖

The match-filename file specifies the file that contains the patterns to match. The file to verify is read from standard input unless the --input-file option is used. 🫶

🎀 OPTIONS 🎀

Options are parsed from the environment variable FILECHECK_OPTS and from the command line. She's got two ears, bestie. 👑

  • -help — Print a summary of command line options.
  • --check-prefixes prefix1,prefix2,... — FileCheck searches the contents of match-filename for patterns to match. By default, these patterns are prefixed with "CHECK:". If you'd like to use a different prefix (e.g. because the same input file is checking multiple different tool or options), the --check-prefixes argument allows you to specify (without the trailing ":") one or more prefixes to match. Multiple prefixes are useful for tests which might change for different run options, but most lines remain the same.

FileCheck does not permit duplicate prefixes, even if one is a check prefix and one is a comment prefix (see --comment-prefixes below). 🚨

  • --check-prefix prefix1,prefix2,... — An alias of --check-prefixes.
  • --comment-prefixes prefix1,prefix2,... — By default, FileCheck ignores any occurrence in match-filename of any check prefix if it is preceded on the same line by "COM:" or "RUN:". See the section The "COM:" directive for usage details.

These default comment prefixes can be overridden by --comment-prefixes if they are not appropriate for your testing environment. However, doing so is not recommended in LLVM's LIT-based test suites, which should be easier to maintain if they all follow a consistent comment style. In that case, consider proposing a change to the default comment prefixes instead. 💜

  • --allow-unused-prefixes — This option controls the behavior when using more than one prefix as specified by --check-prefixes, and some of these prefixes are missing in the test file. If true, this is allowed, if false, FileCheck will report an error, listing the missing prefixes. The default value is false.
  • --input-file filename — File to check (defaults to stdin).
  • --match-full-lines — By default, FileCheck allows matches of anywhere on a line. This option will require all positive matches to cover an entire line. Leading and trailing whitespace is ignored, unless --strict-whitespace is also specified. (Note: negative matches from CHECK-NOT are not affected by this option!)

Passing this option is equivalent to inserting {{^ *}} or {{^}} before, and {{ *$}} or {{$}} after every positive check pattern. 💀

  • --strict-whitespace — By default, FileCheck canonicalizes input horizontal whitespace (spaces and tabs) which causes it to ignore these differences (a space will match a tab). The --strict-whitespace argument disables this behavior. End-of-line sequences are canonicalized to UNIX-style \n in all modes.
  • --ignore-case — By default, FileCheck uses case-sensitive matching. This option causes FileCheck to use case-insensitive matching.
  • --implicit-check-not check-pattern — Adds implicit negative checks for the specified patterns between positive checks. The option allows writing stricter tests without stuffing them with CHECK-NOTs.

For example, "--implicit-check-not warning:" can be useful when testing diagnostic messages from tools that don't have an option similar to clang -verify. With this option FileCheck will verify that input does not contain warnings not covered by any CHECK: patterns. 🦋

  • --dump-input <value> — Dump input to stderr, adding annotations representing currently enabled diagnostics. When there are multiple occurrences of this option, the <value> that appears earliest in the list below has precedence. The default is fail.
    • help — Explain input dump and quit
    • always — Always dump input
    • fail — Dump input on failure
    • never — Never dump input
  • --dump-input-context <N> — In the dump requested by --dump-input, print <N> input lines before and <N> input lines after any lines specified by --dump-input-filter. When there are multiple occurrences of this option, the largest specified <N> has precedence. The default is 5.
  • --dump-input-filter <value> — In the dump requested by --dump-input, print only input lines of kind <value> plus any context specified by --dump-input-context. When there are multiple occurrences of this option, the <value> that appears earliest in the list below has precedence. The default is error when --dump-input=fail, and it's all when --dump-input=always.
    • all — All input lines
    • annotation-full — Input lines with annotations
    • annotation — Input lines with starting points of annotations
    • error — Input lines with starting points of error annotations
  • --enable-var-scope — Enables scope for regex variables.

Variables with names that start with $ are considered global and remain set throughout the file.

All other variables get undefined after each encountered CHECK-LABEL. 🌈

  • -D<VAR=VALUE> — Sets a filecheck pattern variable VAR with value VALUE that can be used in CHECK: lines.
  • -D#<FMT>,<NUMVAR>=<NUMERIC EXPRESSION> — Sets a filecheck numeric variable NUMVAR of matching format FMT to the result of evaluating <NUMERIC EXPRESSION> that can be used in CHECK: lines. See section FileCheck Numeric Variables and Expressions for details on supported numeric expressions.
  • -version — Show the version number of this program.
  • -v — Print good directive pattern matches. However, if -dump-input=fail or -dump-input=always, add those matches as input annotations instead.
  • -vv — Print information helpful in diagnosing internal FileCheck issues, such as discarded overlapping CHECK-DAG: matches, implicit EOF pattern matches, and CHECK-NOT: patterns that do not have matches. Implies -v. However, if -dump-input=fail or -dump-input=always, just add that information as input annotations instead.
  • --allow-deprecated-dag-overlap — Enable overlapping among matches in a group of consecutive CHECK-DAG: directives. This option is deprecated and is only provided for convenience as old tests are migrated to the new non-overlapping CHECK-DAG: implementation. 🚨 deprecated, bestie, so don't get too attached.
  • --allow-empty — Allow checking empty input. By default, empty input is rejected.
  • --color — Use colors in output (autodetected by default).

🦋 EXIT STATUS 🦋

If FileCheck verifies that the file matches the expected contents, it exits with 0. Otherwise, if not, or if an error occurs, it will exit with a non-zero value. Periodt. 💅

🌸 TUTORIAL 🌸

FileCheck is typically used from LLVM regression tests, being invoked on the RUN line of the test. A simple example of using FileCheck from a RUN line looks like this:

; RUN: llvm-as < %s | llc -march=x86-64 | FileCheck %s

This syntax says to pipe the current file ("%s") into llvm-as, pipe that into llc, then pipe the output of llc into FileCheck. This means that FileCheck will be verifying its standard input (the llc output) against the filename argument specified (the original .ll file specified by "%s"). To see how this works, let's look at the rest of the .ll file (after the RUN line):

define void @sub1(i32* %p, i32 %v) {
entry:
; CHECK: sub1:
; CHECK: subl
        %0 = tail call i32 @llvm.atomic.load.sub.i32.p0i32(i32* %p, i32 %v)
        ret void
}

define void @inc4(i64* %p) {
entry:
; CHECK: inc4:
; CHECK: incq
        %0 = tail call i64 @llvm.atomic.load.add.i64.p0i64(i64* %p, i64 1)
        ret void
}

Here you can see some "CHECK:" lines specified in comments. Now you can see how the file is piped into llvm-as, then llc, and the machine code output is what we are verifying. FileCheck checks the machine code output to verify that it matches what the "CHECK:" lines specify. ✨

The syntax of the "CHECK:" lines is very simple: they are fixed strings that must occur in order. FileCheck defaults to ignoring horizontal whitespace differences (e.g. a space is allowed to match a tab) but otherwise, the contents of the "CHECK:" line is required to match some thing in the test file exactly. 🎀

One nice thing about FileCheck (compared to grep) is that it allows merging test cases together into logical groups. For example, because the test above is checking for the "sub1:" and "inc4:" labels, it will not match unless there is a "subl" in between those labels. If it existed somewhere else in the file, that would not count: "grep subl" matches if "subl" exists anywhere in the file. 🫶

✨ The FileCheck -check-prefixes option ✨

The FileCheck -check-prefixes option allows multiple test configurations to be driven from one .ll file. This is useful in many circumstances, for example, testing different architectural variants with llc. Here's a simple example:

; RUN: llvm-as < %s | llc -mtriple=i686-apple-darwin9 -mattr=sse41 \
; RUN:              | FileCheck %s -check-prefixes=X32
; RUN: llvm-as < %s | llc -mtriple=x86_64-apple-darwin9 -mattr=sse41 \
; RUN:              | FileCheck %s -check-prefixes=X64

define <4 x i32> @pinsrd_1(i32 %s, <4 x i32> %tmp) nounwind {
        %tmp1 = insertelement <4 x i32>; %tmp, i32 %s, i32 1
        ret <4 x i32> %tmp1
; X32: pinsrd_1:
; X32:    pinsrd $1, 4(%esp), %xmm0

; X64: pinsrd_1:
; X64:    pinsrd $1, %edi, %xmm0
}

In this case, we're testing that we get the expected code generation with both 32-bit and 64-bit code generation. periodt. 💖

💬 the COM: directive

Okay so sometimes you want to yeet a FileCheck directive out of existence but without actually deleting it? Or you wanna write a comment that name-drops a directive, like, casually? That's literally what COM: is for, bestie. Say you've got this:

; X32: pinsrd_1:
; X32:    pinsrd $1, 4(%esp), %xmm0

; COM: FIXME: X64 isn't working correctly yet for this part of codegen, but
; COM: X64 will have something similar to X32:
; COM:
; COM:   X64: pinsrd_1:
; COM:   X64:    pinsrd $1, %edi, %xmm0

Without COM:, you'd have to do some cursed combo of rewording plus directive syntax mangling to stop FileCheck from clocking those commented X32: and X64: occurrences as real directives. And on top of that — FileCheck diagnostics have been proposed that might complain about those X64 occurrences up there that don't have the trailing : because they look like directive typos. 😭 Dodging all of that is so tedious for a test author, and mangling directive syntax makes the whole point of your test code unclear. COM: avoids all these problems. slay.

A few important usage notes, don't skip these fr:

  • COM: sitting inside another directive's pattern does not comment out the rest of the pattern. Example time:
; X32: pinsrd $1, 4(%esp), %xmm0 COM: This is part of the X32 pattern!

If you need to temporarily comment out part of a directive's pattern, move it to another line. Why? Because FileCheck parses COM: the exact same way as any other directive: only the first directive on the line gets recognized as a directive. periodt.

  • For the sake of LIT, FileCheck treats RUN: just like COM:. If that's not it for your test environment, see --comment-prefixes.
  • FileCheck does not recognize COM, RUN, or any user-defined comment prefix as a comment directive if it's combined with one of the usual check directive suffixes, such as -NEXT: or -NOT:, which we get into below. FileCheck treats that combo as plain text instead. If it needs to act as a comment directive for your test environment, define it as such with --comment-prefixes.

➡️ the CHECK-NEXT: directive

Sometimes you wanna match lines and verify the matches happen on exactly consecutive lines — no other lines sneaking in between. That's your CHECK: + CHECK-NEXT: combo. Got a custom check prefix? Just use <PREFIX>-NEXT:. This totally works how you'd expect:

define void @t2(<2 x double>* %r, <2 x double>* %A, double %B) {
     %tmp3 = load <2 x double>* %A, align 16
     %tmp7 = insertelement <2 x double> undef, double %B, i32 0
     %tmp9 = shufflevector <2 x double> %tmp3,
                            <2 x double> %tmp7,
                            <2 x i32> < i32 0, i32 2 >
     store <2 x double> %tmp9, <2 x double>* %r, align 16
     ret void

; CHECK:          t2:
; CHECK:             movl    8(%esp), %eax
; CHECK-NEXT:        movapd  (%eax), %xmm0
; CHECK-NEXT:        movhpd  12(%esp), %xmm0
; CHECK-NEXT:        movl    4(%esp), %eax
; CHECK-NEXT:        movapd  %xmm0, (%eax)
; CHECK-NEXT:        ret
}

CHECK-NEXT: directives reject the input unless there's exactly one newline between it and the previous directive. And heads up: a CHECK-NEXT: cannot be the first directive in a file. 💀

🌸 the CHECK-SAME: directive

Sometimes you wanna match lines and verify the matches happen on the same line as the previous match. Use CHECK: + CHECK-SAME: for that. Custom check prefix? <PREFIX>-SAME:.

And ooh — CHECK-SAME: is particularly powerful when you pair it with CHECK-NOT: (described below). Obsessed.

Like, this works how you'd expect:

!0 = !DILocation(line: 5, scope: !1, inlinedAt: !2)

; CHECK:       !DILocation(line: 5,
; CHECK-NOT:               column:
; CHECK-SAME:              scope: ![[SCOPE:[0-9]+]]

CHECK-SAME: directives reject the input if there are any newlines between it and the previous directive.

CHECK-SAME: also hits different for avoiding matchers for fields that literally don't matter. Say you're writing a test for a tool that spits out this:

Name: foo
Field1: ...
Field2: ...
Field3: ...
Value: 1

Name: bar
Field1: ...
Field2: ...
Field3: ...
Value: 2

Name: baz
Field1: ...
Field2: ...
Field3: ...
Value: 1

You wanna verify foo has the value 1, so first instinct:

CHECK: Name: foo
CHECK: Value: 1{{$}}

But bestie… that's a bad test 😭. If the value for foo changes, the test still passes, because that CHECK: Value: 1 line would just match the value from baz. Sneaky. To fix it you could add CHECK-NEXT matchers for every FieldN: line, but that's verbose and needs updating the second Field4 shows up. A way more succinct flex with the CHECK-SAME: matcher:

CHECK:      Name: foo
CHECK:      Value:
CHECK-SAME:        {{ 1$}}

This verifies that the next time Value: appears in the output, it has the value 1. chef's kiss.

Note: a CHECK-SAME: cannot be the first directive in a file.

🫥 the CHECK-EMPTY: directive

Need to check that the next line has nothing on it — not even whitespace? That's CHECK-EMPTY:'s whole thing.

declare void @foo()

declare void @bar()
; CHECK: foo
; CHECK-EMPTY:
; CHECK-NEXT: bar

Just like CHECK-NEXT:, the directive fails if there's more than one newline before it finds the next blank line, and it cannot be the first directive in a file.

🚫 the CHECK-NOT: directive

CHECK-NOT: verifies that a string doesn't occur between two matches (or before the first match, or after the last match). So if you wanna prove a load got removed by some transformation, a test like this does the job:

define i8 @coerce_offset0(i32 %V, i32* %P) {
  store i32 %V, i32* %P

  %P2 = bitcast i32* %P to i8*
  %P3 = getelementptr i8* %P2, i32 2

  %A = load i8* %P3
  ret i8 %A
; CHECK: @coerce_offset0
; CHECK-NOT: load
; CHECK: ret i8
}

🔢 the CHECK-COUNT: directive

Gotta match the same pattern over and over and over? You can just repeat a plain CHECK: as many times as you need — but if that looks too boring, use a counted check: CHECK-COUNT-<num>:, where <num> is a positive decimal number. It matches the pattern exactly <num> times — no more, no less. Custom check prefix gang, use <PREFIX>-COUNT-<num>: for the same effect. Simple example:

Loop at depth 1
Loop at depth 1
Loop at depth 1
Loop at depth 1
  Loop at depth 2
    Loop at depth 3

; CHECK-COUNT-6: Loop at depth {{[0-9]+}}
; CHECK-NOT:     Loop at depth {{[0-9]+}}

🕸️ the CHECK-DAG: directive

If you need to match strings that don't occur in strictly sequential order, CHECK-DAG: is your girl. It verifies them between two matches (or before the first match, or after the last match). Fun fact: clang emits vtable globals in reverse order, and with CHECK-DAG: we can keep the checks in the natural order. icon behaviour:

// RUN: %clang_cc1 %s -emit-llvm -o - | FileCheck %s

struct Foo { virtual void method(); };
Foo f;  // emit vtable
// CHECK-DAG: @_ZTV3Foo =

struct Bar { virtual void method(); };
Bar b;
// CHECK-DAG: @_ZTV3Bar =

You can mix CHECK-NOT: directives with CHECK-DAG: directives to exclude strings between the surrounding CHECK-DAG: directives. But as a result, those surrounding CHECK-DAG: directives cannot be reordered — meaning all occurrences matching CHECK-DAG: before CHECK-NOT: must not fall behind occurrences matching CHECK-DAG: after CHECK-NOT:. Like:

; CHECK-DAG: BEFORE
; CHECK-NOT: NOT
; CHECK-DAG: AFTER

This case rejects input strings where BEFORE occurs after AFTER.

Now, with captured variables, CHECK-DAG: can match valid topological orderings of a DAG, with edges going from the definition of a variable to its use. Super useful when your test cases need to match different output sequences from the instruction scheduler, for example:

; CHECK-DAG: add [[REG1:r[0-9]+]], r1, r2
; CHECK-DAG: add [[REG2:r[0-9]+]], r3, r4
; CHECK:     mul r5, [[REG1]], [[REG2]]

In this case, any order of that two add instructions will be allowed. ✨

If you're defining and using variables in the same CHECK-DAG: block, be aware that the definition rule can match after its use.

So for instance, this code below will pass:

; CHECK-DAG: vmov.32 [[REG2:d[0-9]+]][0]
; CHECK-DAG: vmov.32 [[REG2]][1]
vmov.32 d0[1]
vmov.32 d0[0]

While this other code will not:

; CHECK-DAG: vmov.32 [[REG2:d[0-9]+]][0]
; CHECK-DAG: vmov.32 [[REG2]][1]
vmov.32 d1[1]
vmov.32 d0[0]

And yeah, while this can be very useful, it's also dangerous — because with a register sequence you must have a strong order (read before write, copy before use, etc). If the definition your test is looking for doesn't match (because of a bug in the compiler), it may match further away from the use and mask real bugs away. 🙈

In those cases, to enforce the order, use a non-DAG directive between DAG-blocks.

One more thing: a CHECK-DAG: directive skips matches that overlap the matches of any preceding CHECK-DAG: directives in the same CHECK-DAG: block. Not only is this non-overlapping behaviour consistent with other directives, it's also necessary to handle sets of non-unique strings or patterns. For example, these directives look for unordered log entries for two tasks in a parallel program, like the OpenMP runtime:

// CHECK-DAG: [[THREAD_ID:[0-9]+]]: task_begin
// CHECK-DAG: [[THREAD_ID]]: task_end
//
// CHECK-DAG: [[THREAD_ID:[0-9]+]]: task_begin
// CHECK-DAG: [[THREAD_ID]]: task_end

The second pair of directives is guaranteed not to match the same log entries as the first pair — even though the patterns are identical, and even if the text of the log entries is identical, because the thread ID manages to be reused. wild. 💅

The CHECK-LABEL: directive 🎀

Okay so sometimes you've got a file with a bunch of tests split into logical blocks, and one or more CHECK: directives may accidentally succeed by matching lines in a later block. Like, an error will usually eventually get generated, but the check that got flagged as causing the error may not actually have anything to do with the real source of the problem. Messy.

To get better error messages in those cases, you can use the CHECK-LABEL: directive. It's treated identically to a normal CHECK directive except FileCheck makes one extra assumption: a line matched by this directive cannot also be matched by any other check present in match-filename. This is intended for lines containing labels or other unique identifiers. Conceptually, the presence of CHECK-LABEL divides the input stream into separate blocks, each processed independently, which prevents a CHECK: directive in one block from matching a line in another block. If --enable-var-scope is in effect, all local variables are cleared at the beginning of the block.

For example:

define %struct.C* @C_ctor_base(%struct.C* %this, i32 %x) {
entry:
; CHECK-LABEL: C_ctor_base:
; CHECK: mov [[SAVETHIS:r[0-9]+]], r0
; CHECK: bl A_ctor_base
; CHECK: mov r0, [[SAVETHIS]]
  %0 = bitcast %struct.C* %this to %struct.A*
  %call = tail call %struct.A* @A_ctor_base(%struct.A* %0)
  %1 = bitcast %struct.C* %this to %struct.B*
  %call2 = tail call %struct.B* @B_ctor_base(%struct.B* %1, i32 %x)
  ret %struct.C* %this
}

define %struct.D* @D_ctor_base(%struct.D* %this, i32 %x) {
entry:
; CHECK-LABEL: D_ctor_base:

Using CHECK-LABEL: directives here ensures that the three CHECK: directives only accept lines corresponding to the body of the @C_ctor_base function, even if the patterns match lines found later in the file. And bestie, if one of those three CHECK: directives fails, FileCheck will recover by continuing to the next block, which lets multiple test failures get detected in a single invocation. Slay.

There is no requirement that CHECK-LABEL: directives contain strings that correspond to actual syntactic labels in a source or output language: they just have to uniquely match a single line in the file being verified.

CHECK-LABEL: directives cannot contain variable definitions or uses. Periodt.

Directive modifiers ✨

A directive modifier can be appended to a directive by following the directive with {<modifier>}, where the only supported value for <modifier> is LITERAL.

The LITERAL directive modifier can be used to perform a literal match. The modifier results in the directive not recognizing any syntax to perform regex matching, variable capture or any substitutions. This is useful when the text to match would require excessive escaping otherwise. For example, the following will perform literal matches rather than considering these as regular expressions:

Input: [[[10, 20]], [[30, 40]]]
Output %r10: [[10, 20]]
Output %r10: [[30, 40]]

; CHECK{LITERAL}: [[[10, 20]], [[30, 40]]]
; CHECK-DAG{LITERAL}: [[30, 40]]
; CHECK-DAG{LITERAL}: [[10, 20]]

FileCheck Regex Matching Syntax 🦋

All FileCheck directives take a pattern to match. For most uses of FileCheck, fixed string matching is perfectly sufficient. For some things, a more flexible form of matching is desired. To support this, FileCheck lets you specify regular expressions in matching strings, surrounded by double braces: {{yourregex}}. FileCheck implements a POSIX regular expression matcher; it supports Extended POSIX regular expressions (ERE). Because we want to use fixed string matching for a majority of what we do, FileCheck has been designed to support mixing and matching fixed string matching with regular expressions. This lets you write things like this:

; CHECK: movhpd      {{[0-9]+}}(%esp), {{%xmm[0-7]}}

In this case, any offset from the ESP register will be allowed, and any xmm register will be allowed.

Because regular expressions are enclosed with double braces, they are visually distinct, and you don't need to use escape characters within the double braces like you would in C. In the rare case that you want to match double braces explicitly from the input, you can use something ugly like {{[}][}]}} as your pattern. Or if you are using the repetition count syntax, for example [[:xdigit:]]{8} to match exactly 8 hex digits, you would need to add parentheses like this {{([[:xdigit:]]{8})}} to avoid confusion with FileCheck's closing double-brace.

FileCheck String Substitution Blocks 🌸

It is often useful to match a pattern and then verify that it occurs again later in the file. For codegen tests, this can be useful to allow any register, but verify that that register is used consistently later. To do this, FileCheck supports string substitution blocks that allow string variables to be defined and substituted into patterns. Here is a simple example:

; CHECK: test5:
; CHECK:    notw     [[REGISTER:%[a-z]+]]
; CHECK:    andw     {{.*}}[[REGISTER]]

The first check line matches a regex %[a-z]+ and captures it into the string variable REGISTER. The second line verifies that whatever is in REGISTER occurs later in the file after an "andw". FileCheck string substitution blocks are always contained in [[ ]] pairs, and string variable names can be formed with the regex \$[a-zA-Z_][a-zA-Z0-9_]*. If a colon follows the name, then it is a definition of the variable; otherwise, it is a substitution.

FileCheck variables can be defined multiple times, and substitutions always get the latest value. Variables can also be substituted later on the same line they were defined on. For example:

; CHECK: op [[REG:r[0-9]+]], [[REG]]

Can be useful if you want the operands of op to be the same register, and don't care exactly which register it is.

If --enable-var-scope is in effect, variables with names that start with $ are considered to be global. All others variables are local. All local variables get undefined at the beginning of each CHECK-LABEL block. Global variables are not affected by CHECK-LABEL. This makes it easier to ensure that individual tests are not affected by variables set in preceding tests.

FileCheck Numeric Substitution Blocks 🔥

FileCheck also supports numeric substitution blocks that allow defining numeric variables and checking for numeric values that satisfy a numeric expression constraint based on those variables via a numeric substitution. This lets CHECK: directives verify a numeric relation between two numbers, such as the need for consecutive registers to be used.

The syntax to capture a numeric value is [[#%<fmtspec>,<NUMVAR>:]] where:

  • %<fmtspec>, is an optional format specifier to indicate what number format to match and the minimum number of digits to expect.
  • <NUMVAR>: is an optional definition of variable <NUMVAR> from the captured value.

The syntax of <fmtspec> is: #.<precision><conversion specifier> where:

  • # is an optional flag available for hex values (see <conversion specifier> below) which requires the value matched to be prefixed by 0x.
  • .<precision> is an optional printf-style precision specifier in which <precision> indicates the minimum number of digits that the value matched must have, expecting leading zeros if needed.
  • <conversion specifier> is an optional scanf-style conversion specifier to indicate what number format to match (e.g. hex number). Currently accepted format specifiers are %u, %d, %x and %X. If absent, the format specifier defaults to %u.

For example:

; CHECK: mov r[[#REG:]], 0x[[#%.8X,ADDR:]]

would match mov r5, 0x0000FEFE and set REG to the value 5 and ADDR to the value 0xFEFE. Note that due to the precision it would fail to match mov r5, 0xFEFE.

As a result of the numeric variable definition being optional, it is possible to only check that a numeric value is present in a given format. This can be useful when the value itself is not useful, for instance:

; CHECK-NOT: mov r0, r[[#]]

to check that a value is synthesized rather than moved around.

The syntax of a numeric substitution is [[#%<fmtspec>, <constraint> <expr>]] where:

  • <fmtspec> is the same format specifier as for defining a variable but in this context indicating how a numeric expression value should be matched against. If absent, both components of the format specifier are inferred from the matching format of the numeric variable(s) used by the expression constraint if any, and defaults to %u if no numeric variable is used, denoting that the value should be unsigned with no leading zeros. In case of conflict between format specifiers of several numeric variables, the conversion specifier becomes mandatory but the precision specifier remains optional.
  • <constraint> is the constraint describing how the value to match must relate to the value of the numeric expression. The only currently accepted constraint is == for an exact match and is the default if <constraint> is not provided. No matching constraint must be specified when the <expr> is empty.
  • <expr> is an expression. An expression is in turn recursively defined as:
    • a numeric operand, or
    • an expression followed by an operator and a numeric operand.

A numeric operand is a previously defined numeric variable, an integer literal, or a function. Spaces are accepted before, after and between any of these elements. Numeric operands have 64-bit precision. Overflow and underflow are rejected. There is no support for operator precedence, but parentheses can be used to change the evaluation order.

The supported operators are:

  • + - Returns the sum of its two operands.
  • - - Returns the difference of its two operands.

The syntax of a function call is <name>(<arguments>) where:

  • name is a predefined string literal. Accepted values are:
    • add - Returns the sum of its two operands.
    • div - Returns the quotient of its two operands.
    • max - Returns the largest of its two operands.
    • min - Returns the smallest of its two operands.
    • mul - Returns the product of its two operands.
    • sub - Returns the difference of its two operands.
  • <arguments> is a comma separated list of expressions.

For example:

; CHECK: load r[[#REG:]], [r0]
; CHECK: load r[[#REG+1]], [r1]
; CHECK: Loading from 0x[[#%x,ADDR:]]
; CHECK-SAME: to 0x[[#ADDR + 7]]

The above example would match the text:

load r5, [r0]
load r6, [r1]
Loading from 0xa0463440 to 0xa0463447

but would not match the text:

Bestie. When the constraint is actually a numeric constraint, FileCheck starts actually doing math before it moves on, and if the numbers don't line up it drags the whole match down with it:

          load r5, [r0]
          load r7, [r1]
          Loading from 0xa0463440 to 0xa0463443

Because 7 is literally not equal to 5 + 1, and a0463443 is not a0463440 + 7. The math wasn't mathing. 💀

But wait, there's more (there's always more) — a numeric variable can also be defined to the result of a numeric expression. In that case FileCheck checks the numeric expression constraint, and if it verifies, the variable gets assigned the value. Obsessed. The unified syntax for both checking a numeric expression and capturing its value into a numeric variable is:

[[#%<fmtspec>,<NUMVAR>: <constraint> <expr>]]

with every element exactly as described before. You can use this to make your testcase way more self-describing by using variables instead of raw values:

; CHECK: mov r[[#REG_OFFSET:]], 0x[[#%X,FIELD_OFFSET:12]]
; CHECK-NEXT: load r[[#]], [r[[#REG_BASE:]], r[[#REG_OFFSET]]]

which would match:

mov r4, 0xC
load r6, [r5, r4]

so iconic of it, honestly. ✨

The --enable-var-scope option has the exact same effect on numeric variables as it does on string variables. Same vibe, different variable type.

Important note, bestie, do not skip this one: in its current implementation, an expression cannot use a numeric variable defined earlier in the same CHECK directive. That's just how it is rn. 😭

🔮 FileCheck Pseudo Numeric Variables 🔮

Sometimes you need to verify output that contains line numbers of the match file itself — like when you're testing compiler diagnostics. But that introduces a certain fragility into the match file structure, because "CHECK:" lines contain absolute line numbers in the same file, which have to be updated every single time line numbers change from adding or deleting text. So fragile. So high-maintenance. 🙈

To support this case, FileCheck expressions understand the @LINE pseudo numeric variable, which evaluates to the line number of the CHECK pattern where it's found. Main character energy, honestly.

This way match patterns can be put near the relevant test lines and include relative line number references, like this:

// CHECK: test.cpp:[[# @LINE + 4]]:6: error: expected ';' after top level declarator
// CHECK-NEXT: {{^int a}}
// CHECK-NEXT: {{^     \^}}
// CHECK-NEXT: {{^     ;}}
int a

To support legacy uses of @LINE as a special string variable, FileCheck also accepts the following uses of @LINE with string substitution block syntax: [[@LINE]], [[@LINE+<offset>]] and [[@LINE-<offset>]] — with no spaces inside the brackets, where offset is an integer. Little legacy moment, we respect it. 🫶

🌸 Matching Newline Characters 🌸

To match newline characters in regular expressions, you can use the character class [[:space:]]. Like, for example, this pattern:

// CHECK: DW_AT_location [DW_FORM_sec_offset] ([[DLOC:0x[0-9a-f]+]]){{[[:space:]].*}}"intd"

matches output of the form (from llvm-dwarfdump):

DW_AT_location [DW_FORM_sec_offset]   (0x00000233)
DW_AT_name [DW_FORM_strp]  ( .debug_str[0x000000c9] = "intd")

which lets us set the FileCheck variable DLOC to the desired value 0x00000233, extracted from the line immediately preceding "intd". Slay. That's the whole trick and it ate. 🔥

👑 AUTHOR 👑

Maintained by the LLVM Team (https://llvm.org/).

2003-2026, LLVM Project

And that's the whole page, bestie — go forth and CHECK things. 💅 This was the FILECHECK(1) manual page, page 23, dated 2026-10-06. Now go match some patterns, you absolute icon. 🦋🎀💖

📦 where this page comes from