plumbing: transport/ssh, Fix nil pointer dereference caused when an unreachable proxy server is set. Fixes #900

[anandf]

Fixes #900

Root Cause:
variable err is being shadaowed by another variable declaration inside the scope when proxy server url is set. When checking for any error during the Dial operation, the outer scope variable err is used and it would always be nil. Thus the actual error gets ignored and this makes the code execution to proceed with a nil connection which eventually fails with a panic caused due to nil pointer dereference.

Fix:
Change the variable name for the error to dialErr in the outer scope, so that the variable shadowing does not happen.

[anandf]

With this code fix, I am getting the expected error message.

Error:socks connect tcp 127.0.0.1:1080->github.com:22: dial tcp 127.0.0.1:1080: connect: connection refused

Reproducer code:

package main

import (
	"fmt"

	"github.com/go-git/go-git/v5/plumbing/transport"
	"github.com/go-git/go-git/v5/plumbing/transport/client"
	"github.com/go-git/go-git/v5/plumbing/transport/ssh"
)

func main() {
	ep, err := transport.NewEndpoint("git@github.com:anandf/openshift-gitops-microshift.git")
	if err != nil {
		fmt.Println(err)
		return
	}
	ep.Proxy = transport.ProxyOptions{URL: "socks5://127.0.0.1:1080"}
	gitClient, err := client.NewClient(ep)
	if err != nil {
		fmt.Println("Error:" + err.Error())
		return
	}
	auth, err := ssh.NewPublicKeysFromFile("git", "/home/anjoseph/.ssh/id_rsa", "")
	if err != nil {
		fmt.Println("Error:" + err.Error())
		return
	}
	session, err := gitClient.NewUploadPackSession(ep, auth)
	if err != nil {
		fmt.Println("Error:" + err.Error())
		return
	}
	ar, err := session.AdvertisedReferences()
	if err != nil {
		fmt.Println("Error:" + err.Error())
		return
	}
	allRefs, err := ar.AllReferences()
	if err != nil {
		fmt.Println("Error:" + err.Error())
		return
	}
	fmt.Println(allRefs)
	err = session.Close()
	if err != nil {
		fmt.Println("Error:" + err.Error())
		return
	}
}

[pjbgf]

@anandf nice work. Please add a test that reproduces the nil pointer issue, which should fail without the changes introduced by this PR.

[anandf]

@pjbgf Could you please approve the workflows, I want to ensure that the newly added tests are passing. Thanks.

[pjbgf]

@anandf thanks for working on this. 🙇