Add repository fields to all metadata.json files

[adzenith]

There were ten metadata.json files that had no repository set:

$ rg homepage -tjson | wc -l
504
$ rg repository -tjson | wc -l
494

This PR sets all the missing repositories.

[bazel-io]

Hello @fmeum, modules you maintain (bazel_env.bzl, rules_jni, toolchains_musl) have been updated in this PR. Please review the changes.

[bazel-io]

Hello @Vertexwahn, modules you maintain (nasm) have been updated in this PR. Please review the changes.

[bazel-io]

Hello @comius, modules you maintain (rules_cc) have been updated in this PR. Please review the changes.

[bazel-io]

Hello @armandomontanez, @tpudlik, @cramertj, modules you maintain (rules_libusb) have been updated in this PR. Please review the changes.

[bazel-io]

Hello @avdv, @aherrmann, modules you maintain (rules_sh) have been updated in this PR. Please review the changes.

[bazel-io]

Hello @jmillikin, modules you maintain (sqlite3) have been updated in this PR. Please review the changes.

[bazel-io]

Hello @illicitonion, modules you maintain (toolchains_musl) have been updated in this PR. Please review the changes.

[bazel-io]

Hello @bazelbuild/bcr-maintainers, modules without existing maintainers (bzip2, darts-clone) have been updated in this PR. Please review the changes.

[adzenith]

This is the official GitHub mirror. SQLite is developed in a custom VCS so I thought a Git repository link would be more useful

[meteorcloudy]

We should probably add "minItems": 1 at

bazel-central-registry/metadata.schema.json

Lines 35 to 40 in 28b3835

	"repository": {
	"type": "array",
	"items": {
	"description": "repository, typically in the form github:[github org]/[github repo]",
	"type": "string"
	}

to make sure future modules are not missing this attribute?

/cc @alexeagle

[jmillikin]

Modules with a GitHub repository set are required to use source archives hosted on GitHub:

bazel-central-registry/tools/bcr_validation.py

Lines 143 to 173 in 28b3835

	def verify_source_archive_url_match_github_repo(self, module_name, version):
	"""Verify the source archive URL matches the github repo. For now, we only support github repositories check."""
	source_url = self.registry.get_source(module_name, version)["url"]
	source_repositories = self.registry.get_metadata(module_name).get("repository", [])
	matched = not source_repositories
	for source_repository in source_repositories:
	if matched:
	break
	repo_type, repo_path = source_repository.split(":")
	if repo_type == "github":
	parts = urlparse(source_url)
	matched = (
	parts.scheme == "https"
	and parts.netloc == "github.com"
	and os.path.abspath(parts.path).startswith(f"/{repo_path}/")
	)
	elif repo_type == "https":
	repo = urlparse(source_repository)
	parts = urlparse(source_url)
	matched = (
	parts.scheme == repo.scheme
	and parts.netloc == repo.netloc
	and os.path.abspath(parts.path).startswith(f"{repo.path}/")
	)
	if not matched:
	self.report(
	BcrValidationResult.FAILED,
	f"The source URL of {module_name}@{version} ({source_url}) doesn't match any of the module's source repositories {source_repositories}.",
	)
	else:
	self.report(BcrValidationResult.GOOD, "The source URL matches one of the source repositories.")

Wouldn't this change break uploads of modules that have GitHub projects, but host their releases elsewhere?

[meteorcloudy]

Wouldn't this change break uploads of modules that have GitHub projects, but host their releases elsewhere?

http url is also supported as source repo, but yeah, we should make sure they actually match what's being used.

[jmillikin]

I think I'm -1 on this as currently written, after spot-checking a few modules:

• darts-clone has one source archive, under https://storage.googleapis.com/google-code-archive-downloads/, so setting a single github: repository will fail.
• nasm has one source archive, under http://www.nasm.us/pub/nasm/releasebuilds/, so setting a single github: repository will fail.
• Setting the bzip2 repository list to ["git://sourceware.org/git/bzip2.git"] will (if I read the code correctly) cause the source archive validation to always fail, because git: isn't a recognized repository type. You'd need to also have an https: repository in there, presumably https://sourceware.org/pub/bzip2/.
• sqlite3 has source archives under https://sqlite.org/{year}/, so setting a single github: repository will fail.
• rules_libusb has one source archive, under https://storage.googleapis.com/storage/v1/b/pigweed-bazel-tarballs/, so it won't match the repository prefix https://pigweed.googlesource.com/pigweed/rules_libusb.

It seems like there's a conflict of purposes between using repositories as metadata (for linking to upstream source control), and as a validation of archive URLs in BCR tooling.

[Wyverald]

I agree with @jmillikin -- IMO the "repository" field should only be set if we want to check the archive URL, and shouldn't be set just to document the location of the repository (the "homepage" field is perfectly fine for that). However, it could just mean that this PR should be updated to use generic URLs where appropriate instead of github:X/X identifiers as the "repository" field.