winapi GetSystemTimes function does not fill lpIdleTime/lpKernelTime/lpUserTime - winapi

I`m stuck with GetSystemTimes function. There is no any problem in c++ code, but with golang I'm getting three nils instead of expected values
My code:
package main
import (
"log"
"syscall"
"unsafe"
)
var (
modKernel32 = syscall.NewLazyDLL("kernel32.dll")
procGetSystemTimes = modKernel32.NewProc("GetSystemTimes")
)
type SystemTimes struct {
IdleTime *syscall.Filetime
KernelTime *syscall.Filetime
UserTime *syscall.Filetime
}
func main() {
times, err := getSystemTimes()
if err != nil {
log.Fatal(err)
}
log.Printf("idle=%v, kernel=%v, user=%v", times.IdleTime, times.KernelTime, times.UserTime)
}
func getSystemTimes() (times SystemTimes, err error) {
res, _, err1 := procGetSystemTimes.Call(uintptr(unsafe.Pointer(times.IdleTime)), uintptr(unsafe.Pointer(times.KernelTime)), uintptr(unsafe.Pointer(times.UserTime)))
if res != 1 {
return times, err1
}
return
}
It's super simple and very similar to syscall`s GetProcessTimes function from stdlib.
And look, what my program prints:
idle=<nil>, kernel=<nil>, user=<nil>
Why? I expect error or result, why my fields are empty?
Confguration: win8 x64, go1.4.2 windows/amd64

You're passing the address of nil fields to the function.
Either make it point directly to the Filetime struct or initialize all 3 fields before passing them:
type SystemTimes struct {
IdleTime syscall.Filetime
KernelTime syscall.Filetime
UserTime syscall.Filetime
}
.....
res, _, err1 := procGetSystemTimes.Call(uintptr(unsafe.Pointer(&times.IdleTime)), uintptr(unsafe.Pointer(&times.KernelTime)), uintptr(unsafe.Pointer(&times.UserTime)))
Output:
┌─ oneofone#Oa [/tmp]
└──➜ env CC=i686-w64-mingw32-gcc CGO_ENABLED=1 GOOS=windows GOARCH=386 go build -a /tmp/syscall.go; and wine syscall.exe
2015/05/09 22:31:53 idle={1292275456 39}, kernel={1436840864 41}, user={3496565408 2}
playground

Related

How do I check the size of a Go project?

Is there an easy way to check the size of a Golang project? It's not an executable, it's a package that I'm importing in my own project.
You can see how big the library binaries are by looking in the $GOPATH/pkg directory (if $GOPATH is not exported go defaults to $HOME/go).
So to check the size of some of the gorilla http pkgs. Install them first:
$ go get -u github.com/gorilla/mux
$ go get -u github.com/gorilla/securecookie
$ go get -u github.com/gorilla/sessions
The KB binary sizes on my 64-bit MacOS (darwin_amd64):
$ cd $GOPATH/pkg/darwin_amd64/github.com/gorilla/
$ du -k *
284 mux.a
128 securecookie.a
128 sessions.a
EDIT:
Library (package) size is one thing, but how much space that takes up in your executable after the link stage can vary wildly. This is because packages have their own dependencies and with that comes extra baggage, but that baggage may be shared by other packages you import.
An example demonstrates this best:
empty.go:
package main
func main() {}
http.go:
package main
import "net/http"
var _ = http.Serve
func main() {}
mux.go:
package main
import "github.com/gorilla/mux"
var _ = mux.NewRouter
func main() {}
All 3 programs are functionally identical - executing zero user code - but their dependencies differ. The resulting binary sizes in KB:
$ du -k *
1028 empty
5812 http
5832 mux
What does this tell us? The core go pkg net/http adds significant size to our executable. The mux pkg is not large by itself, but it has an import dependency on net/http pkg - hence the significant file size for it too. Yet the delta between mux and http is only 20KB, whereas the listed file size of the mux.a library is 284KB. So we can't simply add the library pkg sizes to determine their true footprint.
Conclusion:
The go linker will strip out a lot of baggage from individual libraries during the build process, but in order to get a true sense of how much extra weight importing certain packages, one has to look at all of the pkg's sub-dependencies as well.
Here is another solution that makes use of https://pkg.go.dev/golang.org/x/tools/go/packages
I took the example provided by the author, and slightly updated it with the demonstration binary available here.
package main
import (
"flag"
"fmt"
"log"
"os"
"sort"
"golang.org/x/tools/go/packages"
)
func main() {
flag.Parse()
// Many tools pass their command-line arguments (after any flags)
// uninterpreted to packages.Load so that it can interpret them
// according to the conventions of the underlying build system.
cfg := &packages.Config{Mode: packages.NeedFiles |
packages.NeedSyntax |
packages.NeedImports,
}
pkgs, err := packages.Load(cfg, flag.Args()...)
if err != nil {
fmt.Fprintf(os.Stderr, "load: %v\n", err)
os.Exit(1)
}
if packages.PrintErrors(pkgs) > 0 {
os.Exit(1)
}
// Print the names of the source files
// for each package listed on the command line.
var size int64
for _, pkg := range pkgs {
for _, file := range pkg.GoFiles {
s, err := os.Stat(file)
if err != nil {
log.Println(err)
continue
}
size += s.Size()
}
}
fmt.Printf("size of %v is %v b\n", pkgs[0].ID, size)
size = 0
for _, pkg := range allPkgs(pkgs) {
for _, file := range pkg.GoFiles {
s, err := os.Stat(file)
if err != nil {
log.Println(err)
continue
}
size += s.Size()
}
}
fmt.Printf("size of %v and deps is %v b\n", pkgs[0].ID, size)
}
func allPkgs(lpkgs []*packages.Package) []*packages.Package {
var all []*packages.Package // postorder
seen := make(map[*packages.Package]bool)
var visit func(*packages.Package)
visit = func(lpkg *packages.Package) {
if !seen[lpkg] {
seen[lpkg] = true
// visit imports
var importPaths []string
for path := range lpkg.Imports {
importPaths = append(importPaths, path)
}
sort.Strings(importPaths) // for determinism
for _, path := range importPaths {
visit(lpkg.Imports[path])
}
all = append(all, lpkg)
}
}
for _, lpkg := range lpkgs {
visit(lpkg)
}
return all
}
You can download all the imported modules with go mod vendor, then count the lines of all the .go files that aren't test files:
package main
import (
"bytes"
"fmt"
"io/fs"
"os"
"os/exec"
"path/filepath"
"strings"
)
func count(mod string) int {
imp := fmt.Sprintf("package main\nimport _ %q", mod)
os.WriteFile("size.go", []byte(imp), os.ModePerm)
exec.Command("go", "mod", "init", "size").Run()
exec.Command("go", "mod", "vendor").Run()
var count int
filepath.WalkDir("vendor", func(s string, d fs.DirEntry, err error) error {
if strings.HasSuffix(s, ".go") && !strings.HasSuffix(s, "_test.go") {
data, err := os.ReadFile(s)
if err != nil {
return err
}
count += bytes.Count(data, []byte{'\n'})
}
return nil
})
return count
}
func main() {
println(count("github.com/klauspost/compress/zstd"))
}

I cannot access err.Err from strconv package

I am probably missing something really simple here:
package main
import (
"fmt"
"strconv"
"reflect"
)
func main() {
s := "abd"
fmt.Println(s)
_, err := strconv.Atoi(s)
if err != nil {
fmt.Println(err)
}
fmt.Println(reflect.TypeOf(err))
fmt.Println(err.Err)
}
I am trying to extract the error itself e.g. ErrSyntax or ErrRange, but I am not able to do so.
After looking at:
https://golang.org/src/strconv/atoi.go?s=3604:3671#L16
I see that err is a pointer to strconv.NumError
15 // A NumError records a failed conversion.
16 type NumError struct {
17 Func string // the failing function (ParseBool, ParseInt, ParseUint, ParseFloat)
18 Num string // the input
19 Err error // the reason the conversion failed (ErrRange, ErrSyntax)
20 }
And Err is the field that holds either ErrRange of ErrSyntax. Therefore, I thought that err.Err would work, but I get:
err.Err undefined (type error has no field or method Err
Err is public, am I missing something with visibility rules?
What am I missing?
Use a type assertion to get the *strconv.NumError value:
if e, ok := err.(*strconv.NumError); ok {
fmt.Println("e.Err", e.Err)
}
playground example

Determining if current process runs in WOW64 or not in Go

With Windows, the official way of guessing if the current 32-bit process is running on a 32 or 64-bit architecture (so on WOW64 or not) is to call the IsWow64Process function from kernel32.dll, and see if it is present (as I understand the doc).
In Go we can call functions exported in dll files with the syscall package, so here is my attempt:
package main
import (
"fmt"
"os"
"syscall"
)
func main() {
dll, err := syscall.LoadDLL("kernel32.dll")
if err != nil {
fmt.Println(err)
}
defer dll.Release()
proc, err := dll.FindProc("IsWow64Process")
if err != nil {
fmt.Println("Proc not found") // not a WOW64 so a 32 bit system?
fmt.Println(err)
}
fmt.Printf("%v\n", proc)
var handle uintptr = uintptr(os.Getpid())
var result uintptr
v, x, y := proc.Call(handle, result)
fmt.Printf("%v %v %v\n", v, x, y)
fmt.Printf("%v\n", result)
}
Unfortunately, testing with or without a WOW64 system displays the same in stdout:
&{0x10ada110 IsWow64Process 2088961457}
0 7 The handle is invalid.
0
What do I do wrong? How to achieve a test to determine if our 32-bit Go program runs on an emulated 32-bit on a 64-bit CPU (WOW64) or on a real 32-bit Windows?
I believe the issue is the handle parameter on your proc.Call. The expected parameter for IsWow64Process is a HANDLE which is not the same as a pid. Which is why it is indicating that the handle is invalid.
The following SO question How to get process handle from process id indicates that you need to call OpenProcess passsing in the pid and it returns the handle.
EDIT: GetCurrentProcess is defined in syscall. So I think you can replace the Getpid call with the following:
handle, err := syscall.GetCurrentProcess()
OK, so here is a working code:
package main
import (
"syscall"
"fmt"
"unsafe"
)
func main() {
dll, err := syscall.LoadDLL("kernel32.dll")
if err != nil {
fmt.Println("Can't load kernel32")
fmt.Println(err)
}
defer dll.Release()
proc, err := dll.FindProc("IsWow64Process")
if err != nil {
fmt.Println("Proc not found")
fmt.Println(err)
}
fmt.Printf("%v\n",proc)
handle, err := syscall.GetCurrentProcess()
if err != nil {
fmt.Println("Handle not found")
fmt.Println(err)
}
fmt.Printf("%v\n",handle)
var result bool
v, x, y := proc.Call(uintptr(handle), uintptr(unsafe.Pointer(&result)))
fmt.Printf("%v %v %v\n",v,x,y)
fmt.Printf("%v\n",result)
}
The result var will be true for a WOW64 system and false for a 32 bit system.
You can also use golang.org/x/sys/windows
package main
import (
"fmt"
"golang.org/x/sys/windows"
)
func main() {
handle := windows.CurrentProcess()
var isWow64 bool
err := windows.IsWow64Process(handle, &isWow64)
if err != nil {
panic(err)
}
fmt.Println(isWow64)
}

Getting count of files in directory using Go

How might I get the count of items returned by io/ioutil.ReadDir()?
I have this code, which works, but I have to think isn't the RightWay(tm) in Go.
package main
import "io/ioutil"
import "fmt"
func main() {
files,_ := ioutil.ReadDir("/Users/dgolliher/Dropbox/INBOX")
var count int
for _, f := range files {
fmt.Println(f.Name())
count++
}
fmt.Println(count)
}
Lines 8-12 seem like way too much to go through to just count the results of ReadDir, but I can't find the correct syntax to get the count without iterating over the range. Help?
Found the answer in http://blog.golang.org/go-slices-usage-and-internals
package main
import "io/ioutil"
import "fmt"
func main() {
files,_ := ioutil.ReadDir("/Users/dgolliher/Dropbox/INBOX")
fmt.Println(len(files))
}
ReadDir returns a list of directory entries sorted by filename, so it is not just files. Here is a little function for those wanting to get a count of files only (and not dirs):
func fileCount(path string) (int, error){
i := 0
files, err := ioutil.ReadDir(path)
if err != nil {
return 0, err
}
for _, file := range files {
if !file.IsDir() {
i++
}
}
return i, nil
}
Starting with Go 1.16 (Feb 2021), a better option is os.ReadDir:
package main
import "os"
func main() {
d, e := os.ReadDir(".")
if e != nil {
panic(e)
}
println(len(d))
}
os.ReadDir returns fs.DirEntry instead of fs.FileInfo, which means that
Size and ModTime methods are omitted, making the process more efficient if
you just need an entry count.
https://golang.org/pkg/os#ReadDir
If you wanna get all files (not recursive) you can use len(files). If you need to just get the files without folders and hidden files just loop over them and increase a counter. And please don’t ignore errors
By looking at the code of ioutil.ReadDir
func ReadDir(dirname string) ([]fs.FileInfo, error) {
f, err := os.Open(dirname)
if err != nil {
return nil, err
}
list, err := f.Readdir(-1)
f.Close()
if err != nil {
return nil, err
}
sort.Slice(list, func(i, j int) bool { return list[i].Name() < list[j].Name() })
return list, nil
}
you would realize that it calls os.File.Readdir() then sorts the files.
In case of counting it, you don't need to sort, so you are better off calling os.File.Readdir() directly.
You can simply copy and paste this function then remove the sort.
But I did find out that f.Readdirnames(-1) is much faster than f.Readdir(-1).
Running time is almost half for /usr/bin/ with 2808 items (16ms vs 35ms).
So to summerize it in an example:
package main
import (
"fmt"
"os"
)
func main() {
f, err := os.Open(os.Args[1])
if err != nil {
panic(err)
}
list, err := f.Readdirnames(-1)
f.Close()
if err != nil {
panic(err)
}
fmt.Println(len(list))
}

Golang, is there a better way read a file of integers into an array?

I need to read a file of integers into an array. I have it working with this:
package main
import (
"fmt"
"io"
"os"
)
func readFile(filePath string) (numbers []int) {
fd, err := os.Open(filePath)
if err != nil {
panic(fmt.Sprintf("open %s: %v", filePath, err))
}
var line int
for {
_, err := fmt.Fscanf(fd, "%d\n", &line)
if err != nil {
fmt.Println(err)
if err == io.EOF {
return
}
panic(fmt.Sprintf("Scan Failed %s: %v", filePath, err))
}
numbers = append(numbers, line)
}
return
}
func main() {
numbers := readFile("numbers.txt")
fmt.Println(len(numbers))
}
The file numbers.txt is just:
1
2
3
...
ReadFile() seems too long (maybe because of the error handing).
Is there a shorter / more Go idiomatic way to load a file?
Using a bufio.Scanner makes things nice. I've also used an io.Reader rather than taking a filename. Often that's a good technique, since it allows the code to be used on any file-like object and not just a file on disk. Here it's "reading" from a string.
package main
import (
"bufio"
"fmt"
"io"
"strconv"
"strings"
)
// ReadInts reads whitespace-separated ints from r. If there's an error, it
// returns the ints successfully read so far as well as the error value.
func ReadInts(r io.Reader) ([]int, error) {
scanner := bufio.NewScanner(r)
scanner.Split(bufio.ScanWords)
var result []int
for scanner.Scan() {
x, err := strconv.Atoi(scanner.Text())
if err != nil {
return result, err
}
result = append(result, x)
}
return result, scanner.Err()
}
func main() {
tf := "1\n2\n3\n4\n5\n6"
ints, err := ReadInts(strings.NewReader(tf))
fmt.Println(ints, err)
}
I would do it like this:
package main
import (
"fmt"
"io/ioutil"
"strconv"
"strings"
)
// It would be better for such a function to return error, instead of handling
// it on their own.
func readFile(fname string) (nums []int, err error) {
b, err := ioutil.ReadFile(fname)
if err != nil { return nil, err }
lines := strings.Split(string(b), "\n")
// Assign cap to avoid resize on every append.
nums = make([]int, 0, len(lines))
for _, l := range lines {
// Empty line occurs at the end of the file when we use Split.
if len(l) == 0 { continue }
// Atoi better suits the job when we know exactly what we're dealing
// with. Scanf is the more general option.
n, err := strconv.Atoi(l)
if err != nil { return nil, err }
nums = append(nums, n)
}
return nums, nil
}
func main() {
nums, err := readFile("numbers.txt")
if err != nil { panic(err) }
fmt.Println(len(nums))
}
Your solution with fmt.Fscanf is fine. There are certainly a number of other ways to do though, depending on your situation. Mostafa's technique is one I use a lot (although I might allocate the result all at once with make. oops! scratch that. He did.) but for ultimate control you should learn bufio.ReadLine. See go readline -> string for some example code.

Resources