Fix error handling

- build.yaml: don't continue on error (this makes the workflow succeed on failures)
- instead, always deploy in scheduled.yaml
- ignore known failures (only issue a warning)

.github/workflows/build.yaml

@@ -130,7 +130,6 @@ jobs:
     runs-on: ${{ inputs.ARCH  == 'x64' && 'ubuntu-22.04' || inputs.ARCH }}
     timeout-minutes: ${{ inputs.BUILD_TIMEOUT || (inputs.ARCH  == 'x64' && 340 || 2880) }}
     name: build debs
-    continue-on-error: true
     outputs:
       LATEST_PACKAGE: ${{ steps.build.outputs.LATEST_PACKAGE }}
 

.github/workflows/interactive.yaml

@@ -86,6 +86,7 @@ jobs:
 
   stage-2:
     needs: stage-1
+    if: needs.stage-1.result == 'success' || ${{ inputs.CONTINUE_ON_ERROR }}
     uses: ubi-agni/ros-builder-action/.github/workflows/build.yaml@main
     with:
       DEB_DISTRO: ${{ inputs.DEB_DISTRO || vars.DEB_DISTRO }}
@@ -109,6 +110,7 @@ jobs:
 
   stage-3:
     needs: stage-2
+    if: needs.stage-2.result == 'success' || ${{ inputs.CONTINUE_ON_ERROR }}
     uses: ubi-agni/ros-builder-action/.github/workflows/build.yaml@main
     with:
       DEB_DISTRO: ${{ inputs.DEB_DISTRO || vars.DEB_DISTRO }}
@@ -131,7 +133,7 @@ jobs:
       INSTALL_TO_CHROOT: ${{ inputs.INSTALL_TO_CHROOT || vars.INSTALL_TO_CHROOT || false }}
 
   deploy:
-    needs: stage-3
+    needs: [stage-1, stage-2, stage-3]
     if: success() && vars.DEPLOY_URL
     timeout-minutes: 15
     runs-on: ubuntu-latest

.github/workflows/scheduled.yaml

@@ -1,6 +1,7 @@
 name: scheduled
 
 on:
+  workflow_dispatch:
   schedule:
     # 8 PM UTC every day
     - cron: "0 20 * * *"
@@ -30,7 +31,7 @@ jobs:
 
   deploy:
     needs: build
-    if: success() && vars.DEPLOY_URL
+    if: always() && vars.DEPLOY_URL
     timeout-minutes: 15
     runs-on: ubuntu-latest
     concurrency:

ros-one.repos

@@ -1060,3 +1060,11 @@ sbuild_options:
   eigenpy:
     amd64: --jobs=2
     arm64: --jobs=1
+
+known_failures:
+  all:
+    - jsk_3rdparty/*
+  jammy:
+  noble:
+    - librealsense2
+    - gazebo_ros_pkgs/gazebo_dev

src/build.sh

@@ -292,6 +292,19 @@ function build_python_pkg {
   ici_label update_repo
 }
 
+function known_to_fail {
+  local pkg_path=$1
+  for key in all "$DEB_DISTRO"; do
+    while IFS= read -r failure; do
+      # shellcheck disable=SC2053
+      if [[ "$pkg_path" == $failure ]]; then
+        return 0
+      fi
+    done < <(yq ".known_failures.${key}[]" "$WS_SOURCE" 2> /dev/null)
+  done
+  return 1
+}
+
 function build_source {
   local old_path=$PWD
   local ws_path="$PWD/ws"
@@ -339,7 +352,9 @@ function build_source {
         *) msg_prefix="unnamed step failed ($exit_code)" ;;
       esac
 
-      if [ "$CONTINUE_ON_ERROR" = false ]; then
+      if known_to_fail "${PKG_FOLDERS[$idx]}"; then
+        gha_warning "$msg_prefix on $pkg_desc."
+      elif [ "$CONTINUE_ON_ERROR" = false ]; then
         # exit with custom error message
         ici_exit "$exit_code" gha_error "$msg_prefix on $pkg_desc."
       else