feat: Add Docker image for auto-indexing support.

.github/workflows/publish-docker.yml

@@ -0,0 +1,52 @@
+name: 'Publish Docker'
+
+on:
+  workflow_dispatch:
+    inputs:
+      tag:
+        description: 'Tag to publish Docker image for (scip-ruby-vM.N.P)'
+  workflow_run:
+    workflows: ['Release']
+    types:
+      - completed
+
+jobs:
+  release-image:
+    runs-on: ubuntu-latest
+    steps:
+      - run: |
+          if [ -z "$TAG" ]; then
+            # From https://stackoverflow.com/questions/10649814/get-last-git-tag-from-a-remote-repo-without-cloning
+            TAG="$(git -c 'versionsort.suffix=-' \
+              ls-remote --exit-code --refs --sort='version:refname' --tags https://github.com/sourcegraph/scip-ruby.git 'scip-ruby-v*' \
+              | tail --lines=1 \
+              | cut --delimiter='/' --fields=3)"
+          fi
+          {
+            echo "TAG=$TAG"
+            echo "PATCH=${TAG/scip-ruby-v/}"
+            echo "MINOR=${PATCH%.*}"
+            echo "MAJOR=${MINOR%.*}"
+          } >> "$GITHUB_ENV"
+        env:
+          TAG: ${{ inputs.tag }}
+      - uses: actions/checkout@v3
+        with:
+          ref: ${{ env.TAG }}
+      - uses: docker/setup-buildx-action@v2
+      - name: Login to DockerHub
+        uses: docker/login-action@v2
+        with:
+          username: ${{ secrets.DOCKER_USERNAME }}
+          password: ${{ secrets.DOCKER_PASSWORD }}
+      - name: Build and push
+        id: docker_build_autoindex
+        uses: docker/build-push-action@v3
+        with:
+          file: Dockerfile.autoindex
+          push: true
+          tags: |
+            sourcegraph/scip-ruby:autoindex
+            sourcegraph/scip-typescript:autoindex-${{ env.PATCH }}
+            sourcegraph/scip-typescript:autoindex-${{ env.MINOR }}
+            sourcegraph/scip-typescript:autoindex-${{ env.MAJOR }}

.github/workflows/release.yml

@@ -1,4 +1,4 @@
-name: Release
+name: Release # This name is used by the publish-docker workflow
 on:
   push:
     tags:

Dockerfile.autoindex

@@ -0,0 +1,20 @@
+FROM --platform=linux/amd64 ruby:2.7.6-alpine3.16@sha256:b014cf3e792d7130daec772241a211c40be009fc7f00e2b728ffe26805649575
+
+# This Docker image is meant for auto-indexing support in Sourcegraph
+# and is not recommended for third-party use.
+
+# gcompat is a glibc-musl compat library (scip-ruby links in glibc)
+# Other deps are to help build C extensions in gems.
+RUN apk add --no-cache bash wget make libstdc++ gcc g++ automake autoconf gcompat
+
+# Use a release binary instead of building from source
+# because release builds are very time-consuming.
+#
+# The release version is verified by tools/scripts/publish-scip-ruby.sh
+RUN wget https://github.com/sourcegraph/scip-ruby/releases/download/scip-ruby-v0.3.0/scip-ruby-x86_64-linux -O /usr/bin/scip-ruby && chmod +x /usr/bin/scip-ruby
+
+COPY scip_indexer/autoindex.sh /usr/bin/scip-ruby-autoindex
+
+RUN chmod +x /usr/bin/scip-ruby-autoindex
+
+CMD ["/bin/sh"]

docs/scip-ruby/CONTRIBUTING.md

@@ -27,6 +27,7 @@ see the [Design Decisions doc][].
 - [Creating PRs](#creating-prs)
 - [Syncing Sorbet upstream](#syncing-sorbet-upstream)
 - [Cutting a release](#cutting-a-release)
+- [Auto-indexing support](#auto-indexing-support)
 - Troubleshooting
   - [Known build issues][]
   - [Known RubyGems related issues](#known-rubygems-related-issues)
@@ -363,6 +364,7 @@ This will correctly use the `scip-ruby/master` branch as the target.
 
 1. Add release notes to the [CHANGELOG](/CHANGELOG.md).
 2. Bump `scip_ruby_version` in `SCIPIndexer.cc`.
+3. Bump the release version in `Dockerfile.autoindex`.
 
 Run the release script:
 
@@ -374,6 +376,36 @@ If there are any errors, fix those and re-run.
 A CI job will be kicked off to trigger a release.
 See the [release workflow](/.github/workflows/release.yml) for details.
 
+## Auto-indexing support
+
+scip-ruby includes has a [Docker image](/Dockerfile.autoindex) for auto-indexing support in Sourcegraph.
+From my testing, this doesn't work under Docker on macOS for some reason
+with a loading error (despite accounting for the glibc-musl mismatch with gcompat).
+It does work on Linux. You can build the Docker image for testing using:
+
+```bash
+docker build . --file Dockerfile.autoindex --platform linux/amd64 --tag scip-ruby-testing
+docker run --platform linux/amd64 --rm -it scip-ruby-testing /bin/bash
+```
+
+Auto-indexing is controlled via a [wrapper script](/scip_indexer/autoindex.sh),
+so that we can handle special cases like RubyGems packages in Sourcegraph,
+without putting that logic into the indexer itself.
+
+You can mimic the workflow of auto-indexing a package by running:
+
+```bash
+PACKAGE=activesupport-7.0.1
+mkdir "$PACKAGE" && cd "$PACKAGE"
+wget "https://rubygems.org/gems/$PACKAGE.gem"
+tar -xf "$PACKAGE.gem"
+tar -xzf data.tar.gz && gunzip metadata.gz && mv metadata rubygems-metadata.yml
+
+scip-ruby-autoindex
+```
+
+For a source repo, cloning the repo and running `scip-ruby-autoindex` should do the trick.
+
 ## Troubleshooting
 
 ### Known build issues

scip_indexer/autoindex.sh

@@ -0,0 +1,85 @@
+#!/usr/bin/env bash
+
+set -euo pipefail
+set -x
+
+while [[ $# -gt 0 ]]; do
+  case $1 in
+    --gem-name)
+      GEM_NAME="$2"
+      shift # past argument
+      shift # past value
+      ;;
+    --gem-version)
+      GEM_VERSION="$2"
+      shift # past argument
+      shift # past value
+      ;;
+    --is-package)
+      IS_PACKAGE_REPO="YES"
+      shift # past argument
+  esac
+done
+
+initialize_sorbet() {
+  gem install "$GEM_NAME" -v "$GEM_VERSION"
+
+  # Assume that current directory is the gem root.
+  # `gem lock` will emit the current gem too, but we don't want that,
+  # as the root already has the source code.
+  #
+  # FIXME: Allow passing in gem source from the outside
+  # Blocked on https://github.com/sourcegraph/sourcegraph/issues/44204
+  gem lock "$GEM_NAME-$GEM_VERSION" \
+   | grep -v "gem '$GEM_NAME'" \
+   | sed -E 's|^gem|source "https://rubygems.org"\ngem|' \
+   > Gemfile
+
+  echo "group :development do
+  gem 'tapioca', require: false
+end" >> Gemfile
+
+  bundle install
+  bundle exec tapioca init
+}
+
+# FIXME: Modify the global gem sources list to use the configured host
+# in Sourcegraph rather than https://rubygems.org
+
+# Created by Sourcegraph to indicate package repos.
+if [ -f "rubygems-metadata.yml" ]; then
+  IS_PACKAGE_REPO="YES"
+fi
+
+if [ -n "${IS_PACKAGE_REPO:-}" ]; then
+  if [ -z "${GEM_NAME:-}" ]; then
+    GEM_NAME="$(grep 'name:' rubygems-metadata.yml | head -n 1 | sed -e 's/name: //')"
+  fi
+  if [ -z "${GEM_VERSION:-}" ]; then
+    GEM_VERSION="$(grep '  version:' rubygems-metadata.yml | head -n 1 | sed -e 's/  version: //')"
+  fi
+  initialize_sorbet
+  set +e
+  scip-ruby . --gem-metadata "$GEM_NAME@$GEM_VERSION"
+  exit 0
+fi
+
+if [ -f "sorbet/config" ]; then
+  set +e
+  if [ -n "${GEM_NAME:-}" ] && [ -n "${GEM_VERSION:-}" ]; then
+    scip-ruby --gem-metadata "$GEM_NAME@$GEM_VERSION"
+  else
+    scip-ruby
+  fi
+  exit 0
+fi
+
+
+set +e
+# Can remove this once the 'default to using .' problem in
+# https://github.com/sourcegraph/scip-ruby/issues/133 is fixed.
+if [ -n "${GEM_NAME:-}" ] && [ -n "${GEM_VERSION:-}" ]; then
+  scip-ruby . --gem-metadata "$GEM_NAME@$GEM_VERSION"
+else
+  scip-ruby .
+fi

tools/scripts/publish-scip-ruby.sh

@@ -21,6 +21,11 @@ if ! grep -q "const char scip_ruby_version\[\] = \"$NEW_VERSION\"" scip_indexer/
   exit 1
 fi
 
+if ! grep -q "download/scip-ruby-$NEW_VERSION" Dockerfile.autoindex; then
+  echo "error: scip-ruby version in Dockerfile is not the latest release version."
+  exit 1
+fi
+
 if ! git diff --quiet; then
   echo "error: Found unstaged changes; aborting."
   exit 1