specific versions. This is necessary to ensure reproducibility of the build.
The pinned versions of the full transitive closure of TensorFlow's dependencies
together with their corresponding hashes are specified in
`requirements_lock_<python version>.txt` files (e.g.
`requirements_lock_3_12.txt` for `Python 3.12`).

To update the lock files, make sure

The main limitation is that to update or delete content of a file inside a ZIP file the entire ZIP file must be replaced.

## How to enable S3 ZIP behavior ?

Ensure to set the following header `x-minio-extract` to `true` in your S3 requests.

## How to access to files inside a ZIP archive

Setting CC is not necessary if the toolchain is built with CC_FOR_TARGET set.

To use the go tool to run individual programs and tests, put $GOROOT/bin into PATH to ensure
the go_ios_$GOARCH_exec wrapper is found. For example, to run the archive/tar tests:

	export PATH=$GOROOT/bin:$PATH
	GOOS=ios GOARCH=amd64 CGO_ENABLED=1 go test archive/tar

		errs, latestMeta, filterDisksByETag, bucket, object, scanMode)

	var erasure Erasure
	if !latestMeta.Deleted && !latestMeta.IsRemote() {
		// Initialize erasure coding
		erasure, err = NewErasure(ctx, latestMeta.Erasure.DataBlocks,
			latestMeta.Erasure.ParityBlocks, latestMeta.Erasure.BlockSize)
		if err != nil {
			return result, err
		}
	}

er/LICENSE).

MinIO cannot make the determination as to whether your application's usage of MinIO is in compliance with the AGPLv3 license requirements. You should instead rely on your own legal counsel or licensing specialists to audit and ensure your application is in compliance with the licenses of MinIO and all other open-source projects with which your application integrates or interacts. We understand that AGPLv3 licensing is complex and nuanced. It is for that reason we strongly encourage...

- Erasure coding used by MinIO is [Reed-Solomon](https://github.com/klauspost/reedsolomon) erasure coding scheme, which has a total shard maximum of 256 i.e 128 data and 128 parity. MinIO design goes beyond this limitation by doing some practical architecture choices.

	Checksum *hash.Checksum
	MD5Hex   string
}

// fanOutPutObject takes an input source reader and fans out multiple PUT operations
// based on the incoming fan-out request, a context cancellation by the caller
// would ensure all fan-out operations are canceled.
func fanOutPutObject(ctx context.Context, bucket string, objectAPI ObjectLayer, fanOutEntries []minio.PutObjectFanOutEntry, fanOutBuf []byte, opts fanOutOptions) ([]ObjectInfo, []error) {

`COMPLETED` only after the replication operation succeeds on each of the targets specified in the replication configuration. If multiple targets are configured to use active-active replication and multi destination replication, the administrator should ensure that the replication features enabled (such as replica metadata sync, delete marker replication etc) are identical to avoid asymmetric state. This is because all replication activity is inherently a one-way operation from source to target, irrespective...

	// allocated slice, but this pattern causes b to escape onto the heap.
	//
	// Instead use the append-make pattern with a nil slice to ensure that
	// we allocate buffers rounded up to the closest size class.
	c := len(b) + n // ensure enough space for n elements
	if c < 2*cap(b) {
		// The growth rate has historically always been 2x. In the future,

currently local to the MinIO server). For this setup, set the `MINIO_BROWSER_REDIRECT_URL` parameter to the publicly/client-accessible endpoint for the MinIO Console. For example `MINIO_BROWSER_REDIRECT_URL=https://console.minio.example.org`. This will ensure that the redirect URL is set to `https://console.minio.example.org/oauth_callback` and the login process should work correctly.

For deployments employing DNS round-robin on a single domain to all the MinIO servers, it is possible that after...