proposal: embed: * pattern should support filenames with single quote

[kazzmir]

Proposal Details

//go:embed x/* currently will not embed filenames in the x subdirectory with a ' single-quote character in their filename, such as

x/tom's favorite file.txt

Single quote is not such a rare character that I think it should be outright rejected by embed. It would be useful if go:embed * could include filenames with a single-quote character in their name.

The globbed filenames are rejected based on the isBadEmbedName in pkg.go:

go/src/cmd/go/internal/load/pkg.go

Line 2271 in e39e965

if err := module.CheckFilePath(name); err != nil {

Where module.CheckFilePath checks for a number of restrictions on the given name:

go/src/cmd/vendor/golang.org/x/mod/module/module.go

Line 495 in e39e965

func CheckFilePath(path string) error {

The commit (930c2c9) that introduced the usage of module.CheckFilePath references CL https://go-review.googlesource.com/c/go/+/290709, which talks a lot about fs.ValidPath. It seems like module.CheckFilePath was used to avoid certain bad windows filenames (COM, PRN, ...), but module.CheckFilePath is fairly restrictive in that it accepts a restricted set of ascii characters:

go/src/cmd/vendor/golang.org/x/mod/module/module.go

Line 285 in e39e965

const allowed = "!#$%&()+,-.=@[]^_{}~ "

This proposal could be resolved by adding single quote ' to that allowed variable, although that would impact other code that uses the golang.org/x/mod/module module. Perhaps a solution would be a combination of fs.ValidPath and something that checks the bad windows names?

func isBadEmbedName(name string) bool {
  if !fs.ValidName(name) {
   return true
  }
  if module.IsBadWindowsName(name) {
    return true
  }
  ...
}

[gabyhelp]

Related Issues

• import/path: embed failure with files with double quotes, special characters #65896 (closed)
• embed: Files starting with - (dash) are not embedded (undocumented behavior) #45447 (closed)
• go:embed directive path.Match patterns will fail any given valid pattern with "syntax error in pattern" if any parent directory contains '[' or ']' characters which are valid folder names characters in windows #68908 (closed)
• proposal: embed: allow embeding files in a parent directory #58519 (closed)
• > Consider the following code: #58526 (closed)
• proposal: embed: treat `/` as module's root in `go:embed` directive #58525 (closed)
• cmd/go,x/mod/module: missing exclusion for COM0, LPT0 for embedded file names #66625 (closed)
• embed: Files starting with : (double dots) are not embedded (undocumented behavior) #46199 (closed)

Related Code Changes

• io/fs: allow backslash in ValidPath, reject in os.DirFS.Open
• cmd/go: reject embedded files that can't be packed into modules

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

[seankhliao]

embed can't do anything until #67562 is resolved.