(aws-codecommit): Repository code initialization fails when empty files exist within the contents

[codeJack]

What is the problem?

When attempting to initialize a CodeCommit repository with some code contents the stack fails with a Code supplied is not a valid .zip archive error if, among the contents, one or more empty files exist (eg. __init__.py).

This behavior applies to any of the supported Code sources : from_asset, from_directory, from_zip_file.

Reproduction Steps

# Example with from_asset
dirname = os.path.dirname(__file__)
repository_code = Asset(
    scope=self,
    id="RepositoryCodeAsset",
    exclude=[".vscode", "__pycache__"],
    path=os.path.join(dirname, "code"),
)

aws_codecommit.Repository(
    scope=self,
    id="CodeCommitRepository",
    repository_name="repository_name",
    code=aws_codecommit.Code.from_asset(asset=repository_code),
)

The code folder shall contain one or more empty files.

What did you expect to happen?

The stack should have been deployed successfully, the code commit repository "repository_name" should have been created and initiated with the contents from the code folder.

Also, the error message could have been more precise : the zip archive is perfectly valid, the issue (assuming there is one) is with its contents.

What actually happened?

cdk deploy fails with a Code supplied is not a valid .zip archive error on the CodeCommit resource.

CDK CLI Version

2.8.0 (build 8a5eb49)

Framework Version

aws-cdk-lib==2.8.0

Node.js Version

v16.10.0

OS

MacOS 11.6.3

Language

Python

Language Version

Python(3.9.10)

Other information

No response

[skinny85]

Hey @codeJack,

thanks a lot for opening the issue. Confirming I was able to reproduce it. Fascinating! I'm curious whether this is limited to CodeCommit, or is this some bigger issue in how we handle Assets in the CDK.

[skinny85]

Interestingly, I just tried it with a Lambda function whose code includes an empty file, and everything worked without a problem!

I wonder if this is some error in how the CodeCommit API handles ZIPs that include empty files...?

[skinny85]

This seems related: jupyterlab/jupyterlab-git#286

[skinny85]

Hmm. I just tried with a manually-created ZIP file that contains just an empty file (not a CDK asset), and it was created successfully. So it does seem like this is something weird that the CDK Assets do.

[jmejco]

repository_code = s3_assets.Asset(
    scope=self,
    id="RepositoryCodeAsset",
    exclude=[".vscode", "__pycache__"],
    path=os.path.join(dirname, "code"),
    bundling=BundlingOptions(
        image=DockerImage.from_registry(
            image="public.ecr.aws/docker/library/alpine:latest"
        ),
        command=[
            "sh",
            "-c",
            """
                apk update && apk add zip
                zip -r -j /asset-output/code.zip /asset-input/*
                """,
        ],
        user="root",
    ),
)

It seems to work fine when using BundlingOptions to create the zip manually.

[skinny85]

What the hell is this ZIP sorcery that is making this fail 😄

[skinny85]

Here's the CDK code creating the archive:

aws-cdk/packages/cdk-assets/lib/private/archive.ts

Lines 9 to 49 in 6eb775e

	export function zipDirectory(directory: string, outputFile: string): Promise<void> {
	return new Promise(async (ok, fail) => {
	// The below options are needed to support following symlinks when building zip files:
	// - nodir: This will prevent symlinks themselves from being copied into the zip.
	// - follow: This will follow symlinks and copy the files within.
	const globOptions = {
	dot: true,
	nodir: true,
	follow: true,
	cwd: directory,
	};
	const files = glob.sync('**', globOptions); // The output here is already sorted
	const output = createWriteStream(outputFile);
	const archive = archiver('zip');
	archive.on('warning', fail);
	archive.on('error', fail);
	// archive has been finalized and the output file descriptor has closed, resolve promise
	// this has to be done before calling `finalize` since the events may fire immediately after.
	// see https://www.npmjs.com/package/archiver
	output.once('close', ok);
	archive.pipe(output);
	// Append files serially to ensure file order
	for (const file of files) {
	const fullPath = path.resolve(directory, file);
	const [data, stat] = await Promise.all([fs.readFile(fullPath), fs.stat(fullPath)]);
	archive.append(data, {
	name: file,
	date: new Date('1980-01-01T00:00:00.000Z'), // reset dates to get the same hash for the same content
	mode: stat.mode,
	});
	}
	await archive.finalize();
	});
	}

I don't see anything suspicious there...

[sawalanipawan]

Got same error when the code contents had empty __init__.py file.

After running cdk deploy command, CDK first creates the zip file and uploads it to the S3 bucket and then initiates the stack creation.

I tried to reference the ZIP created by CDK in below CFN template and tried deploying the stack using this template:

Resources:
  Repository:
    Type: AWS::CodeCommit::Repository
    Properties:
      RepositoryName: MyRepositoryName
      Code:
        BranchName: main
        S3:
          Bucket: my-cdk-bucket
          Key: 'assets/code-zipped-by-cdk.zip'

The stack creation failed with same Code supplied is not a valid .zip archive error.

Then, I tried to create the ZIP manually and added that zip to a S3 bucket. Then I created a CFN stack using below template:

Resources:
  Repository:
    Type: AWS::CodeCommit::Repository
    Properties:
      RepositoryName: MyRepositoryName
      Code:
        BranchName: main
        S3:
          Bucket: mybucket
          Key: 'mymanuallyzippedcode.zip'

The stack was created successfully.

Now, this means that there is an issue in CDK during creation of ZIP file.

[abdulahad95]

Got the same error, was rather confusing to see a "Code supplied is not a valid .zip archive" when the issue was just an empty file in the directory. If at least the error message could be changed that would be helpful

[Snipervuk]

Legend mate

[mi-laf]

Just got the same error. It took me a while to find the reason, also because the .zip archive uploaded to S3 is actually valid and correctly contains the empty file.

[sdpoueme]

Still not working as of today.

[kaiwensun]

It's been a year since the issue is opened. Still not working.

[kaiwensun]

I deleted the .git folder then the stack creation succeeded. I only tested once, so not sure if the .git folder is always the cause. I am using Java Code.fromDirectory

[lemked]

Encountered the same when trying to create a CodeCommit repository from a small directory with some source code.

    const repo = new codecommit.Repository(this, 'Repository', {
      repositoryName: 'sample-app-repo',
      description: 'Repository for sample app.', // optional property
      code: codecommit.Code.fromDirectory(path.join(__dirname, '../../sample-app/'), 'main')
    });

Using a zip file instead worked fine.

    const repo = new codecommit.Repository(this, 'Repository', {
      repositoryName: 'sample-app-repo',
      description: 'Repository for sample app.', // optional property
      code: codecommit.Code.fromZipFile(path.join(__dirname, '../../sample-app/deployment.zip'), 'main')
    });

[dineshmane]

Any updates on this please?

[RyuuyaS]

Can comment inside init file so it is not empty anymore 😀

[davidhessler]

+1 this behavior seems to also be in Typescript