ci: stop third-party artifact downloads reddening unrelated PRs

.github/actions/setup-cloudxr-sdk/action.yml

@@ -30,6 +30,20 @@ runs:
         echo "CloudXR Runtime SDK version: ${CXR_RUNTIME_SDK_VERSION}"
         echo "CloudXR Web SDK version: ${CXR_WEB_SDK_VERSION}"
 
+    # Both download scripts prefer a tarball already sitting in deps/cloudxr/ over
+    # fetching from NGC, so a cache hit makes the download a no-op. NGC 404s
+    # transiently and reddens otherwise-good runs; this keeps the pipeline off the
+    # network on the common path. Restore only — the save is gated on the downloads
+    # succeeding, so a partial fetch is never cached.
+    - name: Restore CloudXR SDK tarballs
+      id: cache
+      uses: actions/cache/restore@55cc8345863c7cc4c66a329aec7e433d2d1c52a9  # v6.1.0
+      with:
+        path: |
+          deps/cloudxr/*.tar.gz
+          deps/cloudxr/*.tgz
+        key: cloudxr-sdk-${{ runner.os }}-${{ runner.arch }}-rt${{ steps.parse.outputs.runtime_version }}-web${{ steps.parse.outputs.web_version }}
+
     - name: Download CloudXR Runtime SDK
       shell: bash
       env:
@@ -44,3 +58,12 @@ runs:
         CXR_WEB_SDK_VERSION: ${{ steps.parse.outputs.web_version }}
       run: |
         ./scripts/download_cloudxr_sdk.sh
+
+    - name: Save CloudXR SDK tarballs
+      if: steps.cache.outputs.cache-hit != 'true'
+      uses: actions/cache/save@55cc8345863c7cc4c66a329aec7e433d2d1c52a9  # v6.1.0
+      with:
+        path: |
+          deps/cloudxr/*.tar.gz
+          deps/cloudxr/*.tgz
+        key: ${{ steps.cache.outputs.cache-primary-key }}

.github/workflows/build-ubuntu.yml

@@ -110,6 +110,21 @@ jobs:
         ccache --version
         ccache --show-config
 
+    # depthai-core downloads the device/bootloader *.fwp blobs into its build
+    # tree at configure time, once per matrix entry; artifacts.luxonis.com fails
+    # often enough to redden unrelated PRs. Its DownloadAndChecksum skips any
+    # file already on disk, so a restored resources/ makes the fetch a no-op.
+    # Keyed on the DEPTHAI_COMMIT pin, which selects the firmware version.
+    # Split restore/save on purpose: the combined action also saves on a failed
+    # job, and CMake's file(DOWNLOAD) leaves a truncated file behind on error —
+    # that would be cached and then pass the "already downloaded" check forever.
+    - name: Restore DepthAI device resources
+      id: depthai-resources
+      uses: actions/cache/restore@55cc8345863c7cc4c66a329aec7e433d2d1c52a9  # v6.1.0
+      with:
+        path: ${{ env.BUILD_DIR }}/_deps/depthai-build/resources
+        key: depthai-resources-${{ runner.os }}-${{ runner.arch }}-${{ hashFiles('cmake/DepthAIVcpkgManifest.cmake') }}
+
     - name: Configure CMake
       run: |
         coverage_args=()
@@ -134,6 +149,14 @@ jobs:
           -DCMAKE_TOOLCHAIN_FILE=${{ env.VCPKG_ROOT }}/scripts/buildsystems/vcpkg.cmake \
           "${coverage_args[@]}"
 
+    # Configure succeeded, so every *.fwp downloaded and checksummed cleanly.
+    - name: Save DepthAI device resources
+      if: steps.depthai-resources.outputs.cache-hit != 'true'
+      uses: actions/cache/save@55cc8345863c7cc4c66a329aec7e433d2d1c52a9  # v6.1.0
+      with:
+        path: ${{ env.BUILD_DIR }}/_deps/depthai-build/resources
+        key: ${{ steps.depthai-resources.outputs.cache-primary-key }}
+
     - name: Build
       run: cmake --build "${BUILD_DIR}" --parallel 4
 

scripts/download_cloudxr_runtime_sdk.sh

@@ -180,34 +180,60 @@ install_from_public_ngc() {
 # Main
 # -----------------------------------------------------------------------------
 
+# One listed -> unlisted sweep over the public NGC resources.
+install_from_ngc() {
+    local resource visibility
+
+    if [[ "$CXR_RUNTIME_SDK_VERSION" == *-rc* ]]; then
+        resource="cloudxr-runtime-for-isaac-teleop"
+        visibility="unlisted"
+    else
+        resource="cloudxr-runtime"
+        visibility="listed"
+    fi
+
+    echo "Trying ${visibility} public NGC..."
+    if install_from_public_ngc "$resource" "$visibility"; then
+        return 0
+    fi
+
+    if [[ "$visibility" == "listed" ]]; then
+        resource="cloudxr-runtime-for-isaac-teleop"
+        visibility="unlisted"
+        echo "Cannot install from listed public NGC, trying unlisted public NGC..."
+        # Do not combine a partial listed download with the unlisted SDK bundle.
+        rm -f "$CXR_DEPLOYMENT_DIR/$SDK_FILE" "$CXR_DEPLOYMENT_DIR/$EXP_SDK_FILE"
+        if install_from_public_ngc "$resource" "$visibility"; then
+            return 0
+        fi
+    fi
+
+    return 1
+}
+
 # Prefer local tarball if present; otherwise use NGC
 if install_from_local_tarball; then
     exit 0
 fi
 
-if [[ "$CXR_RUNTIME_SDK_VERSION" == *-rc* ]]; then
-    NGC_RESOURCE="cloudxr-runtime-for-isaac-teleop"
-    NGC_VISIBILITY="unlisted"
-else
-    NGC_RESOURCE="cloudxr-runtime"
-    NGC_VISIBILITY="listed"
-fi
-
-echo "Cannot install from local tarball, trying ${NGC_VISIBILITY} public NGC..."
-if install_from_public_ngc "$NGC_RESOURCE" "$NGC_VISIBILITY"; then
-    exit 0
-fi
+echo "Cannot install from local tarball, trying public NGC..."
 
-if [[ "$NGC_VISIBILITY" == "listed" ]]; then
-    NGC_RESOURCE="cloudxr-runtime-for-isaac-teleop"
-    NGC_VISIBILITY="unlisted"
-    echo "Cannot install from listed public NGC, trying unlisted public NGC..."
-    # Do not combine a partial listed download with the unlisted SDK bundle.
-    rm -f "$CXR_DEPLOYMENT_DIR/$SDK_FILE" "$CXR_DEPLOYMENT_DIR/$EXP_SDK_FILE"
-    if install_from_public_ngc "$NGC_RESOURCE" "$NGC_VISIBILITY"; then
+# NGC 404s transiently, and curl's --retry never covers 4xx. Retry the whole
+# sweep rather than each request: a 404 on an individual remote name is how
+# download_ngc_first_match discovers the published name, so per-request retry
+# would cost a delay on every successful run.
+NGC_ATTEMPTS=3
+NGC_RETRY_DELAY_S=5
+for (( attempt = 1; attempt <= NGC_ATTEMPTS; attempt++ )); do
+    if install_from_ngc; then
         exit 0
     fi
-fi
+    if (( attempt < NGC_ATTEMPTS )); then
+        echo -e "${YELLOW}NGC download failed (attempt ${attempt}/${NGC_ATTEMPTS}); retrying in ${NGC_RETRY_DELAY_S}s...${NC}"
+        rm -f "$CXR_DEPLOYMENT_DIR/$SDK_FILE" "$CXR_DEPLOYMENT_DIR/$EXP_SDK_FILE"
+        sleep "$NGC_RETRY_DELAY_S"
+    fi
+done
 
-echo "Cannot install from ${NGC_VISIBILITY} public NGC, exiting..."
+echo "Cannot install from public NGC after ${NGC_ATTEMPTS} attempts, exiting..."
 exit 1