Can't load local_repository

[jspkay]

Description of the bug:

I'm working on tensorflow and I want to compile it using a customized version of XNNPACK.
To do so, I wanted ot load the external repository as a local one, as indicated in the documentation.

In brief, I cloned the XNNPACK repo in my laptop and I added the following line to the WORKSPACE:

load("@bazel_tools//tools/build_defs/repo:local.bzl", "local_repository")

Nevertheless, when I try to compile with bazel build, I have the following error message:

ERROR: error loading package '': cannot load '@bazel_tools//tools/build_defs/repo:local.bzl': no such file

I am a noob with bazel, and compilation in general... Sorry if the question is out of place. Do you have any suggestion?

Which category does this issue belong to?

Configurability

What's the simplest, easiest way to reproduce this bug? Please provide a minimal example if possible.

No response

Which operating system are you running Bazel on?

Debian 11

What is the output of bazel info release?

release 5.3.0

If bazel info release returns development version or (@non-git), tell us how you built Bazel.

No response

What's the output of git remote get-url origin; git rev-parse HEAD ?

https://github.com/tensorflow/tensorflow.git
a3e2c692c18649329c4210cf8df2487d2028e267

If this is a regression, please try to identify the Bazel commit where the bug was introduced with bazelisk --bisect.

No response

Have you found anything relevant by searching the web?

I couldn't find anything anywhere, nor with a google search, nor on the issues of the repos.
I though it could be a version problem, but the docs don't say anything regarding it...

Any other information, logs, or outputs that you want to share?

No response

[fmeum]

Bazel 5.3.0 doesn't provide a Starlark implementation of local_repository yet. You can refer to it without a load in your WORKSPACE file. The docs page lacks the version indicator, which is definitely a bug.

[jspkay]

I am sorry to ask this question... But how would I do that then ?
How do I refer to it in my WORKSPACE ?

I tried putting it in the dependencies of my target but it did not work.
I figured bazel need to know that that path contains another repository. Am I wrong?

[fmeum]

@meteorcloudy Do you know whether tensorflow supports being loaded as a dep in this way?

[Wyverald]

This is not a bug in Bazel; please take this to the TensorFlow user forums.