From 26c987c38dc0dc1252a9aded251bebdc35bccffa Mon Sep 17 00:00:00 2001
From: 3x3xX3N0N <300752012+3x3xX3N0N@users.noreply.github.com>
Date: Mon, 3 Aug 2026 18:32:25 -0400
Subject: [PATCH] fix(install): heal the updater's self-owned-marker refusal
 loop on Windows
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

An updater binary spawns 'hermes update' while holding the update marker
with its own PID. A checkout that predates the HERMES_UPDATE_HANDOFF_PID
env fix (8c76fe19f) and the ancestor-pid fallback runs its pre-pull
update_lock.py, reads that marker as a live foreign update, and exits 2
— and the updater deliberately skips its retry for exit 2, so the
refusal loops forever: the update being refused is the one that ships
the fix, and the failure screen's Retry re-enters the same state.

Detect the case with a raw marker read (live_marker_owner deliberately
maps self-ownership to None, so it cannot answer this), drop our own
claim, and retry the child once with the marker absent. The guard
re-removes on Drop (idempotent) and the desktop is already gone at this
point, so nothing races the brief marker-free window.

Fixes #75788

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---
 .../src-tauri/src/update.rs                   | 52 +++++++++++++++++++
 1 file changed, 52 insertions(+)

diff --git a/apps/bootstrap-installer/src-tauri/src/update.rs b/apps/bootstrap-installer/src-tauri/src/update.rs
index df2f0a1bc8..ee11cca232 100644
--- a/apps/bootstrap-installer/src-tauri/src/update.rs
+++ b/apps/bootstrap-installer/src-tauri/src/update.rs
@@ -156,6 +156,24 @@ fn live_marker_owner(path: &Path) -> Option<MarkerOwner> {
     Some(MarkerOwner { pid, age_secs })
 }
 
+/// True when the on-disk marker names THIS process as its owner.
+///
+/// `live_marker_owner` cannot answer this: it deliberately maps
+/// self-ownership to `None` (the #74761 pre-write adoption). The exit-2
+/// self-heal below needs the raw fact, because a `hermes update` child that
+/// refuses over OUR marker is a handoff-recognition failure, not a real
+/// concurrent update.
+fn marker_owned_by_self(path: &Path) -> bool {
+    std::fs::read_to_string(path)
+        .ok()
+        .and_then(|raw| {
+            raw.lines()
+                .next()
+                .and_then(|line| line.trim().parse::<u32>().ok())
+        })
+        == Some(std::process::id())
+}
+
 /// True when a process with `pid` currently exists.
 #[cfg(windows)]
 fn pid_is_alive(pid: u32) -> bool {
@@ -425,6 +443,40 @@ async fn run_update(app: AppHandle) -> Result<()> {
         )
         .await?;
     }
+
+    // Self-owned-marker heal (#75788). Exit 2 means the child refused over a
+    // live update marker with a foreign owner. When that "foreign" owner is
+    // THIS process, the child simply failed to recognize the handoff — a
+    // checkout predating the HERMES_UPDATE_HANDOFF_PID env fix (8c76fe19f)
+    // and the ancestor-pid fallback runs its pre-pull update_lock.py, reads
+    // our marker, and exits 2 every time. The refusal loop is unbreakable
+    // from the user's side because the update being refused is the one that
+    // ships the fix. The marker exists to serialize updates and this process
+    // IS the update: drop our claim and retry once with the marker absent.
+    // The guard re-removes on Drop (idempotent), and the desktop is already
+    // gone at this point, so nothing races the brief marker-free window.
+    if update.exit_code == Some(UPDATE_EXIT_CONCURRENT)
+        && marker_owned_by_self(&crate::paths::update_in_progress_marker())
+    {
+        emit_log(
+            &app,
+            Some("update"),
+            LogStream::Stdout,
+            "[update] child refused over this updater's own marker (stale \
+             checkout without handoff recognition); clearing the claim and \
+             retrying once…",
+        );
+        _update_marker.complete();
+        update = run_streamed(
+            &app,
+            &hermes,
+            &update_args,
+            &install_root,
+            &child_env,
+            Some("update"),
+        )
+        .await?;
+    }
     let update_ms = started.elapsed().as_millis() as u64;
 
     match update.exit_code {