feat: update bazel central registry publish workflow (#23465) (#23913)

Includes pre-built protoc integrity hashes in the release artifact, in a new file `bazel/private/prebuilt_tool_integrity.bzl`

tested on my fork of protobuf: https://github.com/alexeagle/protobuf/releases/tag/v0.1000.10

Next steps:
- create toolchain definitions that fetch the prebuilt protoc from github releases
- change logic in Bazel setup to use the prebuilt toolchain when desired
- create a token from @protobuf-team-bot account and add to GHA secrets so the pull request to BCR opens

Closes #23465

COPYBARA_INTEGRATE_REVIEW=#23465 from protocolbuffers:alexeagle/bazel-protoc-prebuilt 05314db
PiperOrigin-RevId: 817227717

Co-authored-by: Alex Eagle <alex@aspect.dev>

Author: zhangskz

.bcr/source.template.json

@@ -1,5 +1,5 @@
 {
   "integrity": "**leave this alone**",
   "strip_prefix": "{REPO}-{VERSION}",
-  "url": "https://github.com/{OWNER}/{REPO}/releases/download/{TAG}/{REPO}-{VERSION}.zip"
+  "url": "https://github.com/{OWNER}/{REPO}/releases/download/{TAG}/{REPO}-{VERSION}.tar.gz"
 }

.gitattributes

@@ -0,0 +1,4 @@
+# Configuration for 'git archive'
+# see https://git-scm.com/docs/git-archive/2.40.0#ATTRIBUTES
+# Don't include compatibility folder in the distribution artifact, just to reduce size
+compatibility/ export-ignore

.github/workflows/publish_to_bcr.yaml

@@ -0,0 +1,36 @@
+# Publish new releases to Bazel Central Registry.
+name: Publish to BCR
+on:
+  # Run the publish workflow after a successful release
+  # Will be triggered from the release.yaml workflow
+  workflow_call:
+    inputs:
+      tag_name:
+        required: true
+        type: string
+    secrets:
+      # This token should be owned by https://github.com/protobuf-team-bot
+      BCR_PUBLISH_TOKEN:
+        required: true
+  # In case of problems, let release engineers retry by manually dispatching
+  # the workflow from the GitHub UI
+  workflow_dispatch:
+    inputs:
+      tag_name:
+        description: git tag being released
+        required: true
+        type: string
+jobs:
+  publish:
+    uses: bazel-contrib/publish-to-bcr/.github/workflows/publish.yaml@v1.0.0
+    with:
+      tag_name: ${{ inputs.tag_name }}
+      # GitHub repository which is a fork of the upstream where the Pull Request will be opened.
+      registry_fork: protocolbuffers/bazel-central-registry
+    permissions:
+      attestations: write
+      contents: write
+      id-token: write
+    secrets:
+      # Necessary to push to the BCR fork, and to open a pull request against a registry
+      publish_token: ${{ secrets.BCR_PUBLISH_TOKEN }}

.github/workflows/release_bazel_module.yaml

@@ -0,0 +1,33 @@
+# Prepare a release specifically for Bazel users, including a pre-built protoc.
+name: Bazel Release
+on:
+  # Can be triggered from the GitHub Actions ui, using the "Run workflow" button on
+  # https://github.com/protocolbuffers/protobuf/actions/workflows/release_bazel_module.yaml
+  # TODO: consider automating the trigger from a new release being created
+  workflow_dispatch:
+    inputs:
+      tag_name:
+        description: git tag that has the protoc release artifact
+        required: true
+        type: string
+permissions:
+  id-token: write
+  attestations: write
+  contents: write
+jobs:
+  release:
+    uses: bazel-contrib/.github/.github/workflows/release_ruleset.yaml@v7.2.3
+    with:
+      release_files: protobuf-*.tar.gz
+      prerelease: false
+      tag_name: ${{ inputs.tag_name }}
+      # The release was already created by Google-internal mechanism,
+      # so there aren't any useful assertions to make here.
+      bazel_test_command: 'true'
+  publish:
+    needs: release
+    uses: ./.github/workflows/publish_to_bcr.yaml
+    with:
+      tag_name: ${{ inputs.tag_name }}
+    secrets:
+      BCR_PUBLISH_TOKEN: ${{ secrets.BCR_PUBLISH_TOKEN }}

.github/workflows/release_prep.sh

@@ -0,0 +1,53 @@
+#!/usr/bin/env bash
+# NB: this file must be named release_prep.sh because the attestation generation doesn't trust user control.
+# see https://github.com/bazel-contrib/.github/blob/v7.2.3/.github/workflows/release_ruleset.yaml#L33-L45
+set -o errexit -o nounset -o pipefail
+
+# Argument provided by reusable workflow caller, see
+# https://github.com/bazel-contrib/.github/blob/v7.2.3/.github/workflows/release_ruleset.yaml#L104
+TAG=$1
+PREFIX="protobuf-${TAG:1}"
+ARCHIVE="$PREFIX.tar.gz"
+ARCHIVE_TMP=$(mktemp)
+INTEGRITY_FILE=${PREFIX}/bazel/private/prebuilt_tool_integrity.bzl
+
+# NB: configuration for 'git archive' is in /.gitattributes
+git archive --format=tar --prefix=${PREFIX}/ ${TAG} > $ARCHIVE_TMP
+############
+# Patch up the archive to have integrity hashes for built binaries that we downloaded in the GHA workflow.
+# Now that we've run `git archive` we are free to pollute the working directory.
+
+# Delete the placeholder file
+tar --file $ARCHIVE_TMP --delete $INTEGRITY_FILE
+
+# Use jq to translate GitHub Releases json into a Starlark object
+filter_releases=$(cat <<'EOF'
+# Read the file assets already present on the release
+reduce .assets[] as $a (
+  # Start with an empty dictionary, and for each asset, add
+  {}; . + {
+    # The format required in starlark, i.e. "release-name": "deadbeef123"
+    ($a.name): ($a.digest | sub("^sha256:"; "")) 
+  }
+)
+EOF
+)
+
+mkdir -p ${PREFIX}/bazel/private
+cat >${INTEGRITY_FILE} <<EOF
+"Generated during release by release_prep.sh"
+
+RELEASED_BINARY_INTEGRITY = $(
+curl -s https://api.github.com/repos/protocolbuffers/protobuf/releases/tags/${TAG} \
+  | jq -f <(echo "$filter_releases")
+)
+EOF
+
+# Append that generated file back into the archive
+tar --file $ARCHIVE_TMP --append ${INTEGRITY_FILE}
+
+# END patch up the archive
+############
+
+gzip < $ARCHIVE_TMP > $ARCHIVE
+SHA=$(shasum -a 256 $ARCHIVE | awk '{print $1}')

bazel/private/prebuilt_tool_integrity.bzl

@@ -0,0 +1,20 @@
+"""Release binary integrity hashes.
+
+This file contents are entirely replaced during release publishing, by .github/workflows/release_prep.sh
+so that the integrity of the prebuilt tools is included in the release artifact.
+
+The checked in content is only here to allow load() statements in the sources to resolve.
+"""
+
+# Create a mapping for every tool name to the hash of /dev/null
+NULLSHA = "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
+RELEASED_BINARY_INTEGRITY = {
+    "-".join([
+        "protoc",
+        os,
+        arch,
+    ]): NULLSHA
+    for [os, arch] in {
+        "linux": ["aarch_64", "x86_64"],
+    }
+}