From 3100e90453762091999a0a410746fd06ec51c381 Mon Sep 17 00:00:00 2001
From: danielhanchen <unslothai@gmail.com>
Date: Tue, 28 Jul 2026 15:14:06 +0000
Subject: [PATCH] Fail the leg when the installer finished instead of being
 killed

The driver set reason=marker-hit before the post-marker sleep and never
rechecked, so a step whose work was already cached could run to completion
inside that beat and still be recorded as an interruption. The landing assertion
tests reason != marker-hit, so a fully completed install passed green having
interrupted nothing. Reproduced with a stub that exits during the delay:
reported marker-hit / killed=true / exit=0 next to "install finished fully".
Set the reason after the sleep, on both drivers.

Also trigger on _studio_deps.py and install_manifest.py, where the two decisions
the probe asserts on are actually implemented.
---
 .github/scripts/interrupt-install.ps1        |  5 ++++-
 .github/scripts/interrupt-install.sh         | 10 +++++++++-
 .github/workflows/interrupted-install-ci.yml |  6 ++++++
 3 files changed, 19 insertions(+), 2 deletions(-)

diff --git a/.github/scripts/interrupt-install.ps1 b/.github/scripts/interrupt-install.ps1
index 038c6a794a..0cdafed048 100644
--- a/.github/scripts/interrupt-install.ps1
+++ b/.github/scripts/interrupt-install.ps1
@@ -65,8 +65,11 @@ for ($i = 0; $i -lt $KillAtSeconds; $i++) {
   if ($Marker) {
     $hit = Select-String -Path $LogPath -Pattern $Marker -SimpleMatch:$false -ErrorAction SilentlyContinue
     if ($hit) {
-      $reason = 'marker-hit'
       Start-Sleep -Seconds $KillAfterMarkerSeconds
+      # The installer can finish inside the delay; recording marker-hit before it
+      # let a COMPLETED install satisfy the landing assertion and probe HEALTHY.
+      if ($proc.HasExited) { $reason = 'exited-during-marker-delay'; break }
+      $reason = 'marker-hit'
       $killed = $true
       break
     }
diff --git a/.github/scripts/interrupt-install.sh b/.github/scripts/interrupt-install.sh
index 9f3fa3a49d..e805b11d90 100755
--- a/.github/scripts/interrupt-install.sh
+++ b/.github/scripts/interrupt-install.sh
@@ -56,10 +56,18 @@ for i in $(seq 1 "$KILL_AT_SECONDS"); do
     break
   fi
   if [ -n "$MARKER" ] && grep -qE "$MARKER" "$LOG" 2>/dev/null; then
-    reason="marker-hit"
     # Let it get a beat into the step, so the kill lands mid-work rather than on the
     # boundary where the step has not started touching the venv yet.
     sleep "${KILL_AFTER_MARKER_SECONDS:-3}"
+    # ...but a late step whose work is already cached can FINISH inside that beat.
+    # Recording marker-hit before the sleep handed the landing assertion a COMPLETED
+    # install: the signal reached no process, the probe read HEALTHY, and the leg
+    # passed green having interrupted nothing. Set the reason after, not before.
+    if ! kill -0 "$PID" 2>/dev/null; then
+      reason="exited-during-marker-delay"
+      break
+    fi
+    reason="marker-hit"
     killed=true
     break
   fi
diff --git a/.github/workflows/interrupted-install-ci.yml b/.github/workflows/interrupted-install-ci.yml
index 7f09174e17..319cf4ba4e 100644
--- a/.github/workflows/interrupted-install-ci.yml
+++ b/.github/workflows/interrupted-install-ci.yml
@@ -31,6 +31,12 @@ on:
       - 'studio/src-tauri/src/preflight.rs'
       - 'studio/src-tauri/src/preflight/**'
       - 'unsloth_cli/commands/studio.py'
+      # studio_install_ok and verify-install, the two decisions the probe asserts
+      # on, are implemented here rather than in commands/studio.py, so a change
+      # that made install_state() accept a missing manifest would otherwise merge
+      # without a single leg running.
+      - 'unsloth_cli/_studio_deps.py'
+      - 'studio/install_manifest.py'
       # `*` never matches `/`, and it is a literal `-install` that follows, so
       # `interrupt*-install*` matches interrupt-install.sh / .ps1 but NOT the
       # underscored probe. List the probe explicitly rather than rely on a glob.