go.mod: regression: go mod init cannot create module with some unicode charcters if package path is provided on the command line

[SamWhited]

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

$ go version
go version devel +c485506b0aae 20190816195457 linux/amd64

Does this issue reproduce with the latest release?

No (1.12.9 works as expected)

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

go env Output

$ go env
GO111MODULE="on"
GOARCH="amd64"
GOBIN="/home/sam/bin"
GOCACHE="/home/sam/.cache/go-build"
GOENV="/home/sam/.config/go/env"
GOEXE=""
GOFLAGS=""
GOHOSTARCH="amd64"
GOHOSTOS="linux"
GONOPROXY="*"
GONOSUMDB="*"
GOOS="linux"
GOPATH="/home/sam/go"
GOPRIVATE="*"
GOPROXY="127.0.0.1"
GOROOT="/home/sam/Projects/go"
GOSUMDB="off"
GOTMPDIR=""
GOTOOLDIR="/home/sam/Projects/go/pkg/tool/linux_amd64"
GCCGO="gccgo"
AR="ar"
CC="gcc"
CXX="g++"
CGO_ENABLED="1"
GOMOD="/home/sam/go/src/mellium.im/communiqué-shiny/go.mod"
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-build510054645=/tmp/go-build -gno-record-gcc-switches"

What did you do?

$ cd $GOPATH
$ mkdir -p mellium.im/communiqué
$ cd mellium.im/communiqué
$ # This works
$ go mod init
go: creating new go.mod: module mellium.im/communiqué
$ cat go.mod 
module mellium.im/communiqué

go 1.13
$ rm go.mod

$ # This does not work
$ go mod init mellium.im/communiqué
go: malformed import path "mellium.im/communiqué": invalid char 'é'
$ go1.12.9 mod init mellium.im/communiqué
go: creating new go.mod: module mellium.im/communiqué

What did you expect to see?

Running go mod init and go mod init mellium.im/communiqué should both produce a valid go.mod file with some form of Unicode normalization applied.

What did you see instead?

They both behave the same in Go 1.12.9, but do not behave the same in Go 1.13.

[bcmills]

This appears to be a (correct) side effect of https://golang.org/cl/182560, which fixed #32644.

The Go module proxy protocol does not provide an encoding for Unicode characters, and the Go language spec does not define a normalization algorithm for import paths. As a result, a module path containing Unicode characters cannot be fetched or cached: it can only exist as a top-level module, in which case the module path doesn't make much difference anyway.

[bcmills]

Duplicate of #29101