Commit 7d77e311 authored by Rob Pike's avatar Rob Pike

gotest: another try at flags.

doc.go contains the details. The short story:
- command line is passed to the binary
- a new flag, -file, is needed to name files
- known flags have the "test." prefix added for convenience.
- gotest-specific flags are trimmed from the command line.

The effect should be that most existing uses are unaffected,
the ability to name files is still present, and it's nicer to use.
The downside is a lot more code in gotest.

Also allow a test to be called just Test.

R=rsc, niemeyer, rog, r2
CC=golang-dev
https://golang.org/cl/4307049
parent 5a7a0742
......@@ -6,6 +6,7 @@ include ../../Make.inc
TARG=gotest
GOFILES=\
flag.go\
gotest.go\
include ../../Make.cmd
......@@ -6,22 +6,22 @@
Gotest is an automated testing tool for Go packages.
Normally a Go package is compiled without its test files. Gotest
is a simple script that recompiles the package along with any files
named *_test.go. Functions in the test sources named TestXXX
(where XXX is any alphanumeric string starting with an upper case
letter) will be run when the binary is executed. Gotest requires
that the package have a standard package Makefile, one that
includes go/src/Make.pkg.
Normally a Go package is compiled without its test files. Gotest is a
tool that recompiles the package whose source in the current
directory, along with any files named *_test.go. Functions in the
test source named TestXXX (where XXX is any alphanumeric string not
starting with a lower case letter) will be run when the binary is
executed. Gotest requires that the package have a standard package
Makefile, one that includes go/src/Make.pkg.
The test functions are run in the order they appear in the source.
They should have signature
They should have the signature,
func TestXXX(t *testing.T) { ... }
Benchmark functions can be written as well; they will be run only
when the -test.bench flag is provided. Benchmarks should have
signature
Benchmark functions can be written as well; they will be run only when
the -test.bench flag is provided. Benchmarks should have the
signature,
func BenchmarkXXX(b *testing.B) { ... }
......@@ -29,39 +29,47 @@ See the documentation of the testing package for more information.
By default, gotest needs no arguments. It compiles all the .go files
in the directory, including tests, and runs the tests. If file names
are given, only those test files are added to the package.
(The non-test files are always compiled.)
are given (with flag -file=test.go, one per extra test source file),
only those test files are added to the package. (The non-test files
are always compiled.)
The package is built in a special subdirectory so it does not
interfere with the non-test installation.
Usage:
gotest [-c] [-x] [testflags...] [pkg_test.go...]
gotest [-file a.go -file b.go ...] [-c] [-x] [args for test binary]
The flags specific to gotest include -x, which prints each subcommand
gotest executes, and -c, which causes gotest to compile the test
binary but not run it. The testflags are passed to the test binary
and are documented below.
The flags specific to gotest are:
-c Compile the test binary but do not run it.
-file a.go Use the tests in the source file a.go instead of *_test.go.
-x Print each subcommand gotest executes.
Everything else on the command line is passed to the test binary.
The resulting test binary, called (for amd64) 6.out, has several flags.
Usage:
6.out [-test.v] [-test.run pattern] [-test.bench pattern] \
[-test.memprofile=prof.out] [-test.memprofilerate=1]
[-test.cpuprofile=cpu.out] \
[-test.memprofile=mem.out] [-test.memprofilerate=1]
The -test.v flag causes the tests to be logged as they run. The
-test.run flag causes only those tests whose names match the regular
expression pattern to be run. By default all tests are run silently.
If all the specified test pass, 6.out prints PASS and exits with a 0
exit code. If any tests fail, it prints FAIL and exits with a
non-zero code. The -test.bench flag is analogous to the -test.run
flag, but applies to benchmarks. No benchmarks run by default.
expression pattern to be run. By default all tests are run silently.
If all specified tests pass, 6.out prints the word PASS and exits with
a 0 exit code. If any tests fail, it prints error details, the word
FAIL, and exits with a non-zero code. The -test.bench flag is
analogous to the -test.run flag, but applies to benchmarks. No
benchmarks run by default.
The -test.cpuprofile flag causes the testing software to write a CPU
profile to the specified file before exiting.
The -test.memprofile flag causes the testing software to write a
memory profile to the specified file when all tests are complete. Use
-test.run or -test.bench to limit the profile to a particular test or
benchmark. The -test.memprofilerate flag enables more precise (and
expensive) profiles by setting runtime.MemProfileRate;
memory profile to the specified file when all tests are complete. The
-test.memprofilerate flag enables more precise (and expensive)
profiles by setting runtime.MemProfileRate; run
godoc runtime MemProfileRate
for details. The defaults are no memory profile and the standard
setting of MemProfileRate. The memory profile records a sampling of
......@@ -71,16 +79,20 @@ the environment variable GOGC=off to disable the garbage collector,
provided the test can run in the available memory without garbage
collection.
The -test.short package tells long-running tests to shorten their
run time. It is off by default but set by all.bash so installations
of the Go tree can do a sanity check but not spend time running the
full test suite.
Use -test.run or -test.bench to limit profiling to a particular test
or benchmark.
The -test.short package tells long-running tests to shorten their run
time. It is off by default but set by all.bash so installations of
the Go tree can do a sanity check but not spend time running
exhaustive tests.
For convenience, each -test.X flag of the test binary is also
available as the flag -X in gotest itself. For instance, the command
gotest -v -test.cpuprofile=prof.out
will compile the test binary and then run it as
6.out -test.v -cpuprofile=prof.out
For convenience, each of these -test.X flags of the test binary is
also available as the flag -X in gotest itself. Flags not listed here
are unaffected. For instance, the command
gotest -x -v -cpuprofile=prof.out -dir=testdata -update -file x_test.go
will compile the test binary using x_test.go and then run it as
6.out -test.v -test.cpuprofile=prof.out -dir=testdata -update
*/
package documentation
// Copyright 2011 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package main
import (
"fmt"
"os"
"strconv"
"strings"
)
// The flag handling part of gotest is large and distracting.
// We can't use the flag package because some of the flags from
// our command line are for us, and some are for 6.out, and
// some are for both.
var usageMessage = `Usage of %s:
-c=false: compile but do not run the test binary
-file=file:
-x=false: print command lines as they are executed
// These flags can be passed with or without a "test." prefix: -v or -test.v.
-bench="": passes -test.bench to test
-cpuprofile="": passes -test.cpuprofile to test
-memprofile="": passes -test.memprofile to test
-memprofilerate=0: passes -test.memprofilerate to test
-run="": passes -test.run to test
-short=false: passes -test.short to test
-v=false: passes -test.v to test
`
// usage prints a usage message and exits.
func usage() {
fmt.Fprintf(os.Stdout, usageMessage, os.Args[0])
os.Exit(2)
}
// flagSpec defines a flag we know about.
type flagSpec struct {
name string
isBool bool
passToTest bool // pass to Test
multiOK bool // OK to have multiple instances
present bool // flag has been seen
}
// flagDefn is the set of flags we process.
var flagDefn = []*flagSpec{
// gotest-local.
&flagSpec{name: "c", isBool: true},
&flagSpec{name: "file", multiOK: true},
&flagSpec{name: "x", isBool: true},
// passed to 6.out, adding a "test." prefix to the name if necessary: -v becomes -test.v.
&flagSpec{name: "bench", passToTest: true},
&flagSpec{name: "cpuprofile", passToTest: true},
&flagSpec{name: "memprofile", passToTest: true},
&flagSpec{name: "memprofilerate", passToTest: true},
&flagSpec{name: "run", passToTest: true},
&flagSpec{name: "short", isBool: true, passToTest: true},
&flagSpec{name: "v", isBool: true, passToTest: true},
}
// flags processes the command line, grabbing -x and -c, rewriting known flags
// to have "test" before them, and reading the command line for the 6.out.
// Unfortunately for us, we need to do our own flag processing because gotest
// grabs some flags but otherwise its command line is just a holding place for
// 6.out's arguments.
func flags() {
for i := 1; i < len(os.Args); i++ {
arg := os.Args[i]
f, value, extraWord := flag(i)
if f == nil {
args = append(args, arg)
continue
}
switch f.name {
case "c":
setBoolFlag(&cFlag, value)
case "x":
setBoolFlag(&xFlag, value)
case "file":
fileNames = append(fileNames, value)
}
if extraWord {
i++
}
if f.passToTest {
args = append(args, "-test."+f.name+"="+value)
}
}
}
// flag sees if argument i is a known flag and returns its definition, value, and whether it consumed an extra word.
func flag(i int) (f *flagSpec, value string, extra bool) {
arg := os.Args[i]
if strings.HasPrefix(arg, "--") { // reduce two minuses to one
arg = arg[1:]
}
if arg == "" || arg[0] != '-' {
return
}
name := arg[1:]
// If there's already "test.", drop it for now.
if strings.HasPrefix(name, "test.") {
name = name[5:]
}
equals := strings.Index(name, "=")
if equals >= 0 {
value = name[equals+1:]
name = name[:equals]
}
for _, f = range flagDefn {
if name == f.name {
// Booleans are special because they have modes -x, -x=true, -x=false.
if f.isBool {
if equals < 0 { // otherwise, it's been set and will be verified in setBoolFlag
value = "true"
} else {
// verify it parses
setBoolFlag(new(bool), value)
}
} else { // Non-booleans must have a value.
extra = equals < 0
if extra {
if i+1 >= len(os.Args) {
usage()
}
value = os.Args[i+1]
}
}
if f.present && !f.multiOK {
usage()
}
f.present = true
return
}
}
f = nil
return
}
// setBoolFlag sets the addressed boolean to the value.
func setBoolFlag(flag *bool, value string) {
x, err := strconv.Atob(value)
if err != nil {
fmt.Fprintf(os.Stderr, "gotest: illegal bool flag value %s\n", value)
usage()
}
*flag = x
}
......@@ -7,7 +7,6 @@ package main
import (
"bufio"
"exec"
"flag"
"fmt"
"go/ast"
"go/parser"
......@@ -23,14 +22,16 @@ import (
// Environment for commands.
var (
XGC []string // 6g -I _test -o _xtest_.6
GC []string // 6g -I _test _testmain.go
GL []string // 6l -L _test _testmain.6
GOARCH string
GOROOT string
GORUN string
O string
env = os.Environ()
XGC []string // 6g -I _test -o _xtest_.6
GC []string // 6g -I _test _testmain.go
GL []string // 6l -L _test _testmain.6
GOARCH string
GOROOT string
GORUN string
O string
args []string // arguments passed to gotest; also passed to the binary
fileNames []string
env = os.Environ()
)
// These strings are created by getTestNames.
......@@ -44,21 +45,10 @@ var (
importPath string
)
// Flags from package "testing" we will forward to 6.out. See documentation there
// or by running "godoc gotest" - details are in ./doc.go.
// Flags for our own purposes. We do our own flag processing.
var (
test_short bool
test_v bool
test_run string
test_memprofile string
test_memprofilerate int
test_cpuprofile string
)
// Flags for our own purposes
var (
xFlag = flag.Bool("x", false, "print command lines as they are executed")
cFlag = flag.Bool("c", false, "compile but do not run the test binary")
cFlag bool
xFlag bool
)
// File represents a file that contains tests.
......@@ -72,7 +62,7 @@ type File struct {
}
func main() {
flag.Parse()
flags()
needMakefile()
setEnvironment()
getTestFileNames()
......@@ -87,29 +77,11 @@ func main() {
writeTestmainGo()
run(GC...)
run(GL...)
if !*cFlag {
runWithTestFlags("./" + O + ".out")
if !cFlag {
runTestWithArgs("./" + O + ".out")
}
}
// init sets up pairs of flags. Each pair contains a flag defined as in testing, and one with
// the "test." prefix missing, for ease of use.
func init() {
flag.BoolVar(&test_short, "test.short", false, "run smaller test suite to save time")
flag.BoolVar(&test_v, "test.v", false, "verbose: print additional output")
flag.StringVar(&test_run, "test.run", "", "regular expression to select tests to run")
flag.StringVar(&test_memprofile, "test.memprofile", "", "write a memory profile to the named file after execution")
flag.IntVar(&test_memprofilerate, "test.memprofilerate", 0, "if >=0, sets runtime.MemProfileRate")
flag.StringVar(&test_cpuprofile, "test.cpuprofile", "", "write a cpu profile to the named file during execution")
// Now the same flags again, but with shorter names that are forwarded.
flag.BoolVar(&test_short, "short", false, "passes -test.short to test")
flag.BoolVar(&test_v, "v", false, "passes -test.v to test")
flag.StringVar(&test_run, "run", "", "passes -test.run to test")
flag.StringVar(&test_memprofile, "memprofile", "", "passes -test.memprofile to test")
flag.IntVar(&test_memprofilerate, "memprofilerate", 0, "passes -test.memprofilerate to test")
flag.StringVar(&test_cpuprofile, "cpuprofile", "", "passes -test.cpuprofile to test")
}
// needMakefile tests that we have a Makefile in this directory.
func needMakefile() {
if _, err := os.Stat("Makefile"); err != nil {
......@@ -175,7 +147,7 @@ func setEnvironment() {
// getTestFileNames gets the set of files we're looking at.
// If gotest has no arguments, it scans the current directory for _test.go files.
func getTestFileNames() {
names := flag.Args()
names := fileNames
if len(names) == 0 {
names = filepath.Glob("[^.]*_test.go")
if len(names) == 0 {
......@@ -241,9 +213,12 @@ func getTestNames() {
// It is a Test (say) if there is a character after Test that is not a lower-case letter.
// We don't want TesticularCancer.
func isTest(name, prefix string) bool {
if !strings.HasPrefix(name, prefix) || len(name) == len(prefix) {
if !strings.HasPrefix(name, prefix) {
return false
}
if len(name) == len(prefix) { // "Test" is ok
return true
}
rune, _ := utf8.DecodeRuneInString(name[len(prefix):])
return !unicode.IsLower(rune)
}
......@@ -264,34 +239,15 @@ func runWithStdout(argv ...string) string {
return s
}
// runWithTestFlags appends any flag settings to the command line before running it.
func runWithTestFlags(argv ...string) {
if test_short {
argv = append(argv, "-test.short")
}
if test_v {
argv = append(argv, "-test.v")
}
if test_run != "" {
argv = append(argv, fmt.Sprintf("-test.run=%s", test_run))
}
if test_memprofile != "" {
argv = append(argv, fmt.Sprintf("-test.memprofile=%s", test_memprofile))
}
if test_memprofilerate > 0 {
argv = append(argv, fmt.Sprintf("-test.memprofilerate=%d", test_memprofilerate))
}
if test_cpuprofile != "" {
argv = append(argv, fmt.Sprintf("-test.cpuprofile=%s", test_cpuprofile))
}
doRun(argv, false)
// runTestWithArgs appends gotest's runs the provided binary with the args passed on the command line.
func runTestWithArgs(binary string) {
doRun(append([]string{binary}, args...), false)
}
// doRun is the general command runner. The flag says whether we want to
// retrieve standard output.
func doRun(argv []string, returnStdout bool) string {
if *xFlag {
if xFlag {
fmt.Printf("gotest: %s\n", strings.Join(argv, " "))
}
var err os.Error
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment