Refactor git source to allow ScanOptions and use source in engine

[mcastorina]

Refactor the Chunks method of the git Source to call out to two helper methods: scanRepos and scanDirs which scans s.conn.Repositories and s.conn.Directories respectively. The only notable change in behavior is that a credential is no longer necessary if there are no s.conn.Repositories to scan.

[mcastorina]

These are the same options that the git source uses to open a git directory:

trufflehog/pkg/sources/git/git.go

Lines 236 to 242 in 8ec5e49

	func RepoFromPath(path string) (*git.Repository, error) {
	options := &git.PlainOpenOptions{
	DetectDotGit: true,
	EnableDotGitCommonDir: true,
	}
	return git.PlainOpenWithOptions(path, options)
	}

[mcastorina]

Using the source will now ignore any error returned from PlainOpenWithOptions though, but we are preparing the repo already in main which should catch any issues:

trufflehog/main.go

Lines 338 to 341 in 8ec5e49

	repoPath, remote, err = git.PrepareRepoSinceCommit(ctx, *gitScanURI, *gitScanSinceCommit)
	if err != nil || repoPath == "" {
	logFatal(err, "error preparing git repo for scanning")
	}

[mcastorina]

I'm not initializing s.scanOptions, so the default value of nil will be used if it's not set. This is okay because the ScanRepo function already catches this:

trufflehog/pkg/sources/git/git.go

Lines 562 to 565 in 8ec5e49

	func (s *Git) ScanRepo(ctx context.Context, repo *git.Repository, repoPath string, scanOptions *ScanOptions, chunksChan chan *sources.Chunk) error {
	if scanOptions == nil {
	scanOptions = NewScanOptions()
	}

[rosecodym]

If I understand this correctly, you refactored the Git source to allow it to be used in a new way. What are the old ways that it's (presumably) still being used?

[mcastorina]

If I understand this correctly, you refactored the Git source to allow it to be used in a new way. What are the old ways that it's (presumably) still being used?

The main difference is that we weren't doing a git scan using a git.Source and this PR changes that. Instead we were preparing the repo manually and using the dependencies of git.Source to do the scanning.

The problem I encountered when changing it was that the git.Source cleans up temporary directories, while that was previously the responsibility of the code that prepared the repository. I added a kludge in to keep that behavior, though maybe a better solution would be for the git.Source to track which directories were prepared by the Source so it can cleanup without having to rely on naming conventions (though I still want to keep the naming conventions).