Make release aars reproducible

.github/workflows/main.yml

@@ -28,7 +28,10 @@ jobs:
 
     - name: Install gomobile
       run: |
-        go install golang.org/x/mobile/cmd/gomobile@latest
+        # Use the golang.org/x/mobile version go.mod already requires (tidy.yml
+        # updates it with each xray-core bump) rather than whatever @latest is
+        # at build time, so rebuilding a tag later uses the same gomobile.
+        go install golang.org/x/mobile/cmd/gomobile@$(go list -m -f '{{.Version}}' golang.org/x/mobile)
         export PATH=$PATH:~/go/bin
 
     - name: Setup Android SDK
@@ -45,6 +48,18 @@ jobs:
           --install "ndk;29.0.14206865"
         echo "ANDROID_NDK_HOME=$ANDROID_HOME/ndk/29.0.14206865" >> $GITHUB_ENV
 
+    # gomobile compiles the aar's Java bindings (classes.jar) with javac, so the
+    # JDK is part of the output. Without this the runner image's default JDK
+    # is used, which changes whenever GitHub updates the image. This is the
+    # Temurin 17.0.20.1+1 that ubuntu-latest defaults to today, spelled as the
+    # Adoptium semver setup-java matches ('17.0.20.1' would only match the
+    # preinstalled copy by accident, and stop matching once the image moves).
+    - name: Setup Java
+      uses: actions/setup-java@v6.0.0
+      with:
+        distribution: 'temurin'
+        java-version: '17.0.20+101'
+
     - name: Build
       run: |
          mkdir -p assets data
@@ -53,6 +68,13 @@ jobs:
          gomobile init
          go mod tidy
          gomobile bind -v -androidapi 24 -trimpath -ldflags='-s -w -buildid= -checklinkname=0' ./
+         # What a rebuild of this release needs beyond the tagged source: the geo
+         # data resolved above, and the exact Go release setup-go picked for
+         # go.mod's `go` line. Published next to the aar.
+         { cat data/geo-assets.lock; echo "GO_VERSION='$(go env GOVERSION)'"; } > libv2ray-build.lock
+         cat libv2ray-build.lock
+         # Checksums of what this run built, to compare a rebuild against
+         sha256sum libv2ray*r
 
     - name: Upload build artifacts
       if: github.event.inputs.release_tag == ''
@@ -61,6 +83,7 @@ jobs:
         name: libv2ray
         path: |
           ${{ github.workspace }}/libv2ray*r
+          ${{ github.workspace }}/libv2ray-build.lock
 
     - name: Upload AndroidLibXrayLite to release
       if: github.event.inputs.release_tag != ''
@@ -69,3 +92,10 @@ jobs:
         file: ./libv2ray*r
         tag: ${{ github.event.inputs.release_tag }}
         file_glob: true
+
+    - name: Upload build lock to release
+      if: github.event.inputs.release_tag != ''
+      uses: svenstaro/upload-release-action@v2
+      with:
+        file: ./libv2ray-build.lock
+        tag: ${{ github.event.inputs.release_tag }}

README.md

@@ -11,3 +11,15 @@
 2. `gomobile init`
 3. `go mod tidy -v`
 4. `gomobile bind -v -androidapi 24 -trimpath -ldflags='-s -w -buildid= -checklinkname=0' ./`
+
+## Reproducing a release
+Each release publishes a `libv2ray-build.lock` next to `libv2ray.aar`, recording
+the inputs its tagged source does not pin by itself: the Go release it was built
+with, and the exact geo data embedded in `assets/`. To rebuild a release's aar
+byte for byte:
+
+1. Check out the release tag, and use the JDK, NDK and Android SDK versions named in `.github/workflows/main.yml`
+2. Install the Go release named by `GO_VERSION` in `libv2ray-build.lock`
+3. `mkdir -p assets data && bash gen_assets.sh pinned libv2ray-build.lock && cp data/*.dat assets/`
+4. `go install golang.org/x/mobile/cmd/gomobile@$(go list -m -f '{{.Version}}' golang.org/x/mobile)`
+5. `gomobile init`, `go mod tidy -v`, then the `gomobile bind` command from step 4 above

gen_assets.sh

@@ -10,30 +10,101 @@ __file="${__dir}/$(basename "${BASH_SOURCE[0]}")"
 __base="$(basename "${__file}" .sh)"
 
 DATADIR="${__dir}/data"
+# Written by every run: each source resolved to an exact release tag or commit,
+# with the SHA-256 of what was downloaded. Published with the aar, it lets a
+# release be rebuilt with the same data via `gen_assets.sh pinned <file>`.
+LOCKFILE="${DATADIR}/geo-assets.lock"
+
+RULES_REPO="Loyalsoldier/v2ray-rules-dat"
+GEOIP_REPO="Loyalsoldier/geoip"
 
 
 # Check for required dependencies
 check_dependencies() {
     command -v jq >/dev/null 2>&1 || { echo >&2 "jq is required but it's not installed. Aborting."; exit 1; }
     command -v go >/dev/null 2>&1 || { echo >&2 "Go is required but it's not installed. Aborting."; exit 1; }
+    command -v git >/dev/null 2>&1 || { echo >&2 "git is required but it's not installed. Aborting."; exit 1; }
 }
 
 
-# Download data function
-download_dat() {
+# The tag that github.com/<repo>/releases/latest currently redirects to
+latest_release_tag() {
+    curl -fsSI "https://github.com/$1/releases/latest" | awk -F/ 'tolower($0) ~ /^location:/ {print $NF}' | tr -d '\r'
+}
+
+# The commit a branch currently points at
+branch_commit() {
+    git ls-remote "https://github.com/$1.git" "refs/heads/$2" | cut -f1
+}
+
+sha256() {
+    if command -v sha256sum >/dev/null 2>&1; then sha256sum "$1"; else shasum -a 256 "$1"; fi | cut -d' ' -f1
+}
+
+# fetch <url> <file> [expected sha256]: download, verify if a checksum is
+# given, and print the file's checksum
+fetch() {
+    curl -fsSL "$1" -o "$2"
+    local sum
+    sum=$(sha256 "$2")
+    if [[ -n "${3:-}" && "$sum" != "$3" ]]; then
+        echo >&2 "Checksum mismatch for $2: expected $3, got $sum"
+        exit 1
+    fi
+    echo "$sum"
+}
+
+# Download from the sources named by RULES_DAT_TAG and GEOIP_RELEASE_COMMIT,
+# verify against any *_SHA256 already set, and record what was fetched
+fetch_dat() {
     if [[ ! -d "$DATADIR" ]]; then
         echo "Downloading failed \"$DATADIR\" does not exists"
         exit 1
     fi
 
-    echo "Downloading geoip.dat..."
-    curl -sL https://github.com/Loyalsoldier/v2ray-rules-dat/releases/latest/download/geoip.dat -o "$DATADIR/geoip.dat"
+    echo "Downloading geoip.dat from $RULES_REPO $RULES_DAT_TAG..."
+    GEOIP_DAT_SHA256=$(fetch "https://github.com/$RULES_REPO/releases/download/$RULES_DAT_TAG/geoip.dat" "$DATADIR/geoip.dat" "${GEOIP_DAT_SHA256:-}")
+
+    echo "Downloading geosite.dat from $RULES_REPO $RULES_DAT_TAG..."
+    GEOSITE_DAT_SHA256=$(fetch "https://github.com/$RULES_REPO/releases/download/$RULES_DAT_TAG/geosite.dat" "$DATADIR/geosite.dat" "${GEOSITE_DAT_SHA256:-}")
+
+    echo "Downloading geoip-only-cn-private.dat from $GEOIP_REPO $GEOIP_RELEASE_COMMIT..."
+    GEOIP_ONLY_CN_PRIVATE_DAT_SHA256=$(fetch "https://raw.githubusercontent.com/$GEOIP_REPO/$GEOIP_RELEASE_COMMIT/geoip-only-cn-private.dat" "$DATADIR/geoip-only-cn-private.dat" "${GEOIP_ONLY_CN_PRIVATE_DAT_SHA256:-}")
 
-    echo "Downloading geosite.dat..."
-    curl -sL https://github.com/Loyalsoldier/v2ray-rules-dat/releases/latest/download/geosite.dat -o "$DATADIR/geosite.dat"
+    cat > "$LOCKFILE" <<EOF
+RULES_DAT_TAG='$RULES_DAT_TAG'
+GEOIP_DAT_SHA256='$GEOIP_DAT_SHA256'
+GEOSITE_DAT_SHA256='$GEOSITE_DAT_SHA256'
+GEOIP_RELEASE_COMMIT='$GEOIP_RELEASE_COMMIT'
+GEOIP_ONLY_CN_PRIVATE_DAT_SHA256='$GEOIP_ONLY_CN_PRIVATE_DAT_SHA256'
+EOF
+    echo "Recorded the exact sources in $LOCKFILE"
+}
 
-    echo "Downloading geoip-only-cn-private.dat..."
-    curl -sL https://raw.githubusercontent.com/Loyalsoldier/geoip/release/geoip-only-cn-private.dat -o "$DATADIR/geoip-only-cn-private.dat"
+# Download data function: the newest data, as before, but resolved to an
+# exact tag and commit first so that what was downloaded can be recorded
+download_dat() {
+    RULES_DAT_TAG=$(latest_release_tag "$RULES_REPO")
+    GEOIP_RELEASE_COMMIT=$(branch_commit "$GEOIP_REPO" release)
+    if [[ -z "$RULES_DAT_TAG" || -z "$GEOIP_RELEASE_COMMIT" ]]; then
+        echo >&2 "Could not resolve the latest geo data sources. Aborting."
+        exit 1
+    fi
+    fetch_dat
+}
+
+# Exactly the data recorded in a lock file, for rebuilding a release
+pinned_dat() {
+    local lock="${1:-}"
+    if [[ -z "$lock" ]]; then
+        echo >&2 "Usage: gen_assets.sh pinned <lock file>"
+        exit 1
+    fi
+    # `.` searches PATH for a bare file name; make sure the given file is read
+    [[ "$lock" == */* ]] || lock="./$lock"
+    # shellcheck source=/dev/null
+    . "$lock"
+    fetch_dat
 }
 
 # Main execution logic
@@ -43,5 +114,6 @@ check_dependencies
 
 case $ACTION in
     "download") download_dat ;;
+    "pinned") pinned_dat "${2:-}" ;;
     *) echo "Invalid action: $ACTION" ; exit 1 ;;
 esac