cmd/compile: incorrect handling of iota in 1.18

[willeccles]

What version of Go are you using (go version)?

$ go version
go version go1.18.1 linux/amd64

Does this issue reproduce with the latest release?

Yes.

What operating system and processor architecture are you using (go env)?

go env Output

$ go env
GO111MODULE=""
GOARCH="amd64"
GOBIN=""
GOCACHE="/home/cactus/.cache/go-build"
GOENV="/home/cactus/.config/go/env"
GOEXE=""
GOEXPERIMENT=""
GOFLAGS=""
GOHOSTARCH="amd64"
GOHOSTOS="linux"
GOINSECURE=""
GOMODCACHE="/home/cactus/go/pkg/mod"
GONOPROXY=""
GONOSUMDB=""
GOOS="linux"
GOPATH="/home/cactus/go"
GOPRIVATE=""
GOPROXY="https://proxy.golang.org,direct"
GOROOT="/usr/lib/go"
GOSUMDB="sum.golang.org"
GOTMPDIR=""
GOTOOLDIR="/usr/lib/go/pkg/tool/linux_amd64"
GOVCS=""
GOVERSION="go1.18.1"
GCCGO="gccgo"
GOAMD64="v1"
AR="ar"
CC="gcc"
CXX="g++"
CGO_ENABLED="1"
GOMOD="/home/cactus/git/golua/go.mod"
GOWORK=""
CGO_CFLAGS="-g -O2"
CGO_CPPFLAGS=""
CGO_CXXFLAGS="-g -O2"
CGO_FFLAGS="-g -O2"
CGO_LDFLAGS="-g -O2"
PKG_CONFIG="pkg-config"
GOGCCFLAGS="-fPIC -m64 -pthread -fmessage-length=0 -fdebug-prefix-map=/tmp/go-build2033115951=/tmp/go-build -gno-record-gcc-switches"

What did you do?

Using golua, 1.18.1 causes panics which did not occur on previous versions of go. I have narrowed down the issue to this source file which functions differently now. Running this program on 1.18.1 yields different results than 1.17.9, but according to the language spec, 1.18.1 is exhibiting incorrect behavior.

What did you expect to see?

In the above program, OpAnd should have a value of 1 (1 + iota<<8 should be 1 + (0)<<8).

What did you see instead?

On go 1.17.9, the value is indeed 1. However, on go 1.18.1, the value is 257, meaning that iota here is incorrectly being set to 1 instead of 0. The language spec says that each constant declaration should reset iota to 0, which is not happening on 1.18.1.

[willeccles]

Original discussion of this issue can be found here: arnodel/golua#87

[ianlancetaylor]

Here is a standalone program showing the problem. This program should not panic. It does not panic when compiled with Go 1.17 or earlier, or with gccgo. It does panic with Go 1.18.

package main

const C1 = 0 + iota<<8

const C2 = 1 + iota<<8

func main() {
	if C1 != 0 {
		panic(C1)
	}
	if C2 != 1 {
		panic(C2)
	}
}

CC @griesemer @mdempsky

[ianlancetaylor]

@gopherbot Please open a backport to 1.18.

This is an invalid compilation of a valid Go file. Earlier releases are unaffected.

[gopherbot]

Backport issue(s) opened: #52441 (for 1.18).

Remember to create the cherry-pick CL(s) as soon as the patch is submitted to master, according to https://go.dev/wiki/MinorReleases.

[zikaeroh]

FWIW I tried to bisect this, and it blamed the commit where -G=3 was enabled by default; re-bisecting between 1.17 and 1.18 with -G=3 fails immediately because the issue appears to predate those releases. Going back further gets into the weeds since it's probably not something that was introduced via some specific commit thanks to the WIP-ness of the compiler at that point.

[griesemer]

No need to bisect. This is due to a completely new type checker, so it won't be useful to pin-point to a single change. I've identified the bug and will have a fix in a little bit.

This is clearly a bad bug; but only manifests itself when using iota outside a grouped constant declaration, twice.

As a temporary work-around, you can change your code to:

// OpOr is a logical or (precedence 0)
const (OpOr Op = 0 + iota<<8)

// OpAnd is a logical and (precedence 1)
const (OpAnd Op = 1 + iota<<8)

(put parentheses around the const declarations).

[gopherbot]

Change https://go.dev/cl/401134 mentions this issue: cmd/compile/internal/types2: use correct value of iota