fix: scope apt regen to dispatched channel (stop cross-channel index clobber) — #3218

.github/workflows/wheels-released.yml

@@ -199,6 +199,11 @@ jobs:
         env:
           GPG_PASSPHRASE: ${{ secrets.WHEELS_REPO_GPG_PASSPHRASE }}
           GPG_KEY_ID: ${{ env.GPG_KEY_ID }}
+          # Scope regen to the dispatched channel only. The sync step above pulls
+          # just pool/<channel>/ from R2, so regenerating the OTHER channel would
+          # scan an empty local pool and clobber its R2 index on upload
+          # (#3218 / #2838). The other channel's dists are left untouched.
+          CHANNELS: ${{ steps.inputs.outputs.channel }}
         run: |
           set -euo pipefail
           chmod +x scripts/regenerate-apt-metadata.sh

scripts/regenerate-apt-metadata.sh

@@ -20,7 +20,17 @@ fi
 
 ARCHITECTURES="amd64 arm64"
 COMPONENTS="main"
-DISTRIBUTIONS="stable bleeding-edge"
+# Per-channel regen (CHANNELS env), default = both for manual full rebuilds.
+#
+# CRITICAL (#3218, recurrence of #2838): the publish workflow only syncs
+# pool/<dispatched-channel>/ from R2, so any channel NOT being published has an
+# empty local pool here. Regenerating it would emit an empty Packages, and the
+# upload step's `find dists` would then clobber that channel's good R2 index.
+# A bleeding-edge snapshot publish was wiping the stable index minutes after
+# every stable release. Scoping to the dispatched channel keeps the other
+# channel's R2 dists untouched. The workflow passes CHANNELS=<channel>; a bare
+# manual run still rebuilds both (only safe when both pools are present locally).
+DISTRIBUTIONS="${CHANNELS:-stable bleeding-edge}"
 
 # apt-ftparchive uses a config file to know where the pool lives. The same
 # config drives both distributions — only the dist-name and the scan path