math/rand: Source.Intn consistently returns 0

[zchenyu]

Go version

go version go1.24.0 linux/amd64

Output of go env in your module/workspace:

AR='ar'
CC='gcc'
CGO_CFLAGS='-O2 -g'
CGO_CPPFLAGS=''
CGO_CXXFLAGS='-O2 -g'
CGO_ENABLED='0'
CGO_FFLAGS='-O2 -g'
CGO_LDFLAGS='-O2 -g'
CXX='g++'
GCCGO='gccgo'
GO111MODULE=''
GOAMD64='v1'
GOARCH='amd64'
GOAUTH='netrc'
GOBIN=''
GOCACHE='/home/zchenyu/.cache/go-build'
GOCACHEPROG=''
GODEBUG=''
GOENV='/home/zchenyu/.config/go/env'
GOEXE=''
GOEXPERIMENT=''
GOFIPS140='off'
GOFLAGS=''
GOGCCFLAGS='-fPIC -m64 -Wl,--no-gc-sections -fmessage-length=0 -ffile-prefix-map=/tmp/go-build814464281=/tmp/go-build -gno-record-gcc-switches'
GOHOSTARCH='amd64'
GOHOSTOS='linux'
GOINSECURE=''
GOMOD='/home/zchenyu/tmp/go/go.mod'
GOMODCACHE='/home/zchenyu/go/pkg/mod'
GONOPROXY='github.com/fw-ai'
GONOSUMDB='github.com/fw-ai'
GOOS='linux'
GOPATH='/home/zchenyu/go'
GOPRIVATE='github.com/fw-ai'
GOPROXY='https://proxy.golang.org,direct'
GOROOT='/home/zchenyu/go/pkg/mod/golang.org/toolchain@v0.0.1-go1.24.0.linux-amd64'
GOSUMDB='sum.golang.org'
GOTELEMETRY='local'
GOTELEMETRYDIR='/home/zchenyu/.config/go/telemetry'
GOTMPDIR=''
GOTOOLCHAIN='auto'
GOTOOLDIR='/home/zchenyu/go/pkg/mod/golang.org/toolchain@v0.0.1-go1.24.0.linux-amd64/pkg/tool/linux_amd64'
GOVCS=''
GOVERSION='go1.24.0'
GOWORK=''
PKG_CONFIG='pkg-config'

What did you do?

Eerily similar to #51015
Our generation code is a little more complex (has prefixes/suffixes) than the original post, hence the strings.Builder, but it's the same in spirit.

We use GenerateID in an RPC service and after several months (years?) of no issue, for a period of a few hours, we were consistently generating the string aaaaaaaa in about ~30% of calls. Other timing info:

• The issue was observed only in one server replica, the others were okay
• Only started several hours/days after the server was started, and it stopped by itself
• Roughly speaking, each RPC calls GenerateID once. Not all calls resulted in the string of aaaaaaaa, only ~30%. The other generated strings looked normal.

I also cannot reproduce the issue in a small program.

const charset = "abcdefghijklmnopqrstuvwxyz0123456789"
var seededRand = rand.New(rand.NewSource(time.Now().UnixNano()))

func GenerateID(length int) string {
	var sb strings.Builder
	sb.Grow(length)
	for i := 0; i < length; i++ {
		sb.WriteByte(charset[seededRand.Intn(len(charset))])
	}
	return sb.String()
}

In any case, it seems seeding is no longer necessary after go1.20, so we will just remove the NewSource and use the default (concurrency safe) one. Hopefully that fixes the problem.

What did you see happen?

Consistently see 0's returned by Intn()

What did you expect to see?

Random numbers

[seankhliao]

like before, I don't see any actionable item here.
Note that you should prefer math/rand/v2 if you're making a change.

[gabyhelp]

Related Issues

• math/rand: After ~1.5 month of valid usage, rand started consistently returning 0 #51015 (closed)
• math/rand (or crypto/rsa): non-deterministic output after go1.11.2 #29290 (closed)
• math/rand should use not use static default seed #45618 (closed)
• fmt.Print affect rand.Intn ? #48461 (closed)
• math/rand: Uses the same random cycle every time #53355 (closed)

_{(Emoji vote if this was helpful or unhelpful; more detailed feedback welcome in this discussion.)}

[zchenyu]

Ha yeah fine to close. Just trying to add some more data points in case this others run into it and if someone decides to look into it. The original ticket was closed so I couldn't comment.

[randall77]

Did you run with the race detector? As mentioned in #51015, that was likely the issue there.
A rand.Source (which is what rand.NewSource returns) is not safe for use by multiple goroutines at once.