Yes, I think that we shouldn't be following path.Clean quite so closely here.

The quick fix is to change

	if len(elem) > 0 {
		elem = append([]string{u.Path}, elem...)
		url.setPath(path.Join(elem...))
	}

to

	if len(elem) > 0 {
		elem = append([]string{u.Path}, elem...)
                p := path.Join(elem...)
                if strings.HasSuffix(elem[len(elem)-1], "/") {
                   p += "/"
                }
		url.setPath(p)
	}

Another approach is to copy and modify path.Join. Doing a copy could save some allocations by working directly in a strings.Builder instead of bouncing between []byte and string.

@gopherbot Please open backport to 1.18

This function is new in 1.18 so we should keep it consistent with this change.

Backport issue(s) opened: #52087 (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.

Change https://go.dev/cl/397256 mentions this issue: net/url: preserve a trailing slash in JoinPath

The function is slated for 1.19. It missed 18.

Whoops. Thanks.

What should url.JoinPath do in these cases:

url.JoinPath("https://example.com/a//")
url.JoinPath("https://example.com/a//", "")

I think the output should be "https://example.com/a//" and "https://example.com/a/" but I can be argued into thinking otherwise.

Personally I think it's OK if JoinPath drops extra trailing slashes. I guess it should be documented, though.