cmd/go: the go command converts uppercase characters to !+ lowercase characters.

[saibing]

Please answer these questions before submitting your issue. Thanks!

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

go version devel +c54bc34 Thu Jul 19 00:23:07 2018 +0000 linux/amd64

Does this issue reproduce with the latest release?

yes

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

baas@SZX1000428520:~/git/go/src/go-langserver [master]$ go env
GOARCH="amd64"
GOBIN=""
GOCACHE="/home/baas/.cache/go-build"
GOEXE=""
GOHOSTARCH="amd64"
GOHOSTOS="linux"
GOOS="linux"
GOPATH="/home/baas/git/go"
GOPROXY="http://10.75.203.36:9090"
GORACE=""
GOROOT="/home/baas/my/go"
GOTMPDIR=""
GOTOOLDIR="/home/baas/my/go/pkg/tool/linux_amd64"
GCCGO="gccgo"
CC="gcc"
CXX="g++"
CGO_ENABLED="1"
GOMOD="/home/baas/git/go/src/go-langserver/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-build071732067=/tmp/go-build -gno-record-gcc-switches"

What did you do?

I am using the go 1.11 moudle function, I developed a go module proxy and used it by setting the environment variable GOPROXY

What did you expect to see?

the go module proxy can work well.

What did you see instead?

• client

$ go get github.com/SimpleMessageNotification/smn-sdk-go@latest
go get github.com/SimpleMessageNotification/smn-sdk-go@latest: cannot find module providing package github.com/SimpleMessageNotification/smn-sdk-go

• server

$ ./vgoproxy
0719 10:48:40.33 go config:
{
   "gopath": "/data1/baas/.gomod",
   "http": [
      "code.huawei.com",
      "rnd-isource.huawei.com",
      "rnd-github.huawei.com"
   ],
   "replace": {},
   "sortKeys": null
}
0719 10:48:40.33 start go proxy server at :9090
0719 10:49:23.565 GET /github.com/!simple!message!notification/smn-sdk-go/@v/list from 100.101.229.133:47174
0719 10:49:23.565 new url /github.com/!simple!message!notification/smn-sdk-go/@v/list
0719 10:49:23.566 go: list version failed: invalid github.com/ import path "github.com/!simple!message!notification/smn-sdk-go"
0719 10:49:23.58 GET /github.com/!simple!message!notification/@v/list from 100.101.229.133:47174
0719 10:49:23.58 new url /github.com/!simple!message!notification/@v/list
0719 10:49:23.58 go: list version failed: invalid github.com/ import path "github.com/!simple!message!notification"
0719 10:49:23.591 GET /github.com/@v/list from 100.101.229.133:47174
0719 10:49:23.591 new url /github.com/@v/list
0719 10:49:24.718 go: list version failed: unrecognized import path "github.com" (parse https://github.com?go-get=1: no go-import meta tags ())

[magical]

This appears to be working as intended:

go/src/cmd/go/internal/module/module.go

Lines 359 to 417 in c54bc34

	// Safe encodings
	//
	// Module paths appear as substrings of file system paths
	// (in the download cache) and of web server URLs in the proxy protocol.
	// In general we cannot rely on file systems to be case-sensitive,
	// nor can we rely on web servers, since they read from file systems.
	// That is, we cannot rely on the file system to keep rsc.io/QUOTE
	// and rsc.io/quote separate. Windows and macOS don't.
	// Instead, we must never require two different casings of a file path.
	// Because we want the download cache to match the proxy protocol,
	// and because we want the proxy protocol to be possible to serve
	// from a tree of static files (which might be stored on a case-insensitive
	// file system), the proxy protocol must never require two different casings
	// of a URL path either.
	//
	// One possibility would be to make the safe encoding be the lowercase
	// hexadecimal encoding of the actual path bytes. This would avoid ever
	// needing different casings of a file path, but it would be fairly illegible
	// to most programmers when those paths appeared in the file system
	// (including in file paths in compiler errors and stack traces)
	// in web server logs, and so on. Instead, we want a safe encoding that
	// leaves most paths unaltered.
	//
	// The safe encoding is this:
	// replace every uppercase letter with an exclamation mark
	// followed by the letter's lowercase equivalent.
	//
	// For example,
	// github.com/Azure/azure-sdk-for-go -> github.com/!azure/azure-sdk-for-go.
	// github.com/GoogleCloudPlatform/cloudsql-proxy -> github.com/!google!cloud!platform/cloudsql-proxy
	// github.com/Sirupsen/logrus -> github.com/!sirupsen/logrus.
	//
	// Import paths that avoid upper-case letters are left unchanged.
	// Note that because import paths are ASCII-only and avoid various
	// problematic punctuation (like : < and >), the safe encoding is also ASCII-only
	// and avoids the same problematic punctuation.
	//
	// Import paths have never allowed exclamation marks, so there is no
	// need to define how to encode a literal !.
	//
	// Although paths are disallowed from using Unicode (see pathOK above),
	// the eventual plan is to allow Unicode letters as well, to assume that
	// file systems and URLs are Unicode-safe (storing UTF-8), and apply
	// the !-for-uppercase convention. Note however that not all runes that
	// are different but case-fold equivalent are an upper/lower pair.
	// For example, U+004B ('K'), U+006B ('k'), and U+212A ('K' for Kelvin)
	// are considered to case-fold to each other. When we do add Unicode
	// letters, we must not assume that upper/lower are the only case-equivalent pairs.
	// Perhaps the Kelvin symbol would be disallowed entirely, for example.
	// Or perhaps it would encode as "!!k", or perhaps as "(212A)".
	//
	// Also, it would be nice to allow Unicode marks as well as letters,
	// but marks include combining marks, and then we must deal not
	// only with case folding but also normalization: both U+00E9 ('é')
	// and U+0065 U+0301 ('e' followed by combining acute accent)
	// look the same on the page and are treated by some file systems
	// as the same path. If we do allow Unicode marks in paths, there
	// must be some kind of normalization to allow only one canonical
	// encoding of any character used in an import path.

[oiooj]

As @magical said, this appears to be working as intended. You need to decode the path before you use it.

go/src/cmd/go/internal/module/module.go

Line 455 in c54bc34

func DecodePath(encoding string) (path string, err error) {

[saibing]

@magical @magical

Thank you very much.

github.com/golang/vgo does not sync the latest code

[magical]

github.com/golang/vgo does not sync the latest code

Correct. Per Russ's recent email to golang-dev,

Work on the module support will now be done exclusively in the main repo. We will automatically export the main repo to the x/vgo repo when we have a snapshot we want to make available to people using "vgo".

https://groups.google.com/forum/#!topic/golang-dev/a5PqQuBljF4