From 85fffd9fb78b68aab1266c540a2aceffb0d298db Mon Sep 17 00:00:00 2001
From: openhands <openhands@all-hands.dev>
Date: Sat, 21 Mar 2026 11:22:07 +0000
Subject: [PATCH] =?UTF-8?q?fix:=20CI=20fix=20counter=20can=20be=20wasted?=
 =?UTF-8?q?=20by=20guard=20hits=20=E2=80=94=20consider=20check=5Fonly=20mo?=
 =?UTF-8?q?de=20(#377)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Orphan and stuck-PR CI-failure paths in dev-poll.sh called
handle_ci_exhaustion without check_only, incrementing the fix counter on
every poll cycle even when guards (session checks, is_blocked) prevented
an actual agent launch. This could exhaust the 3-attempt budget without
any real fix attempts.

Now both paths use the same two-phase pattern as the backlog scan:
1. check_only during the scan (no counter increment)
2. Increment atomically at actual launch time

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---
 dev/dev-poll.sh | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/dev/dev-poll.sh b/dev/dev-poll.sh
index ebf46cd..09387a5 100755
--- a/dev/dev-poll.sh
+++ b/dev/dev-poll.sh
@@ -375,10 +375,14 @@ if [ "$ORPHAN_COUNT" -gt 0 ]; then
         log "issue #${ISSUE_NUM} already has active session ${SESSION_NAME} — skipping"
         exit 0
       fi
-      if handle_ci_exhaustion "$HAS_PR" "$ISSUE_NUM"; then
+      if handle_ci_exhaustion "$HAS_PR" "$ISSUE_NUM" "check_only"; then
         # Fall through to backlog scan instead of exit
         :
       else
+        # Increment at actual launch time (not on guard-hit paths)
+        if handle_ci_exhaustion "$HAS_PR" "$ISSUE_NUM"; then
+          exit 0  # exhausted between check and launch
+        fi
         log "issue #${ISSUE_NUM} PR #${HAS_PR} CI failed — spawning agent to fix (attempt ${CI_FIX_ATTEMPTS}/3)"
         nohup "${SCRIPT_DIR}/dev-agent.sh" "$ISSUE_NUM" >> "$LOGFILE" 2>&1 &
         log "started dev-agent PID $! for issue #${ISSUE_NUM} (CI fix)"
@@ -483,9 +487,13 @@ for i in $(seq 0 $(($(echo "$OPEN_PRS" | jq 'length') - 1))); do
       log "issue #${STUCK_ISSUE} already has active session ${SESSION_NAME} — skipping"
       continue
     fi
-    if handle_ci_exhaustion "$PR_NUM" "$STUCK_ISSUE"; then
+    if handle_ci_exhaustion "$PR_NUM" "$STUCK_ISSUE" "check_only"; then
       continue  # skip this PR, check next stuck PR or fall through to backlog
     fi
+    # Increment at actual launch time (not on guard-hit paths)
+    if handle_ci_exhaustion "$PR_NUM" "$STUCK_ISSUE"; then
+      continue  # exhausted between check and launch
+    fi
     log "PR #${PR_NUM} (issue #${STUCK_ISSUE}) CI failed — fixing (attempt ${CI_FIX_ATTEMPTS}/3)"
     nohup "${SCRIPT_DIR}/dev-agent.sh" "$STUCK_ISSUE" >> "$LOGFILE" 2>&1 &
     log "started dev-agent PID $! for stuck PR #${PR_NUM}"