mergerfs/vendored
Antonio S. Musumeci 81e8f6281a fix data race on _threads between constructor and autoscale workers
The constructor's spawn loop pushed to _threads without holding
_threads_mutex.  A just-spawned autoscale worker could time out on
the empty queue (especially with a small idle_threshold_usecs) and
call _try_remove_self_if_above, which acquires _threads_mutex and
modifies _threads (find + erase).  Concurrent unlocked push_back
and locked erase on the same vector is undefined behavior per the
C++ memory model, even when reserve() prevents reallocation.

The same race applied to the _threads.empty() and _threads.size()
reads after the spawn loop.

Fix: lock _threads_mutex around push_back in the spawn loop and
around the post-loop size check.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-10 22:24:14 -04:00
..
acutest Update vendored/acutest to latest master 2026-03-02 01:09:16 -05:00
boost Update vendored/boost to 1.90.0 2026-03-02 01:09:16 -05:00
CLI11 Update vendored/CLI11 to v2.6.2 2026-03-02 01:09:16 -05:00
fmt Move fmtlib to vendored and upgrade to 12.1.0 2026-03-02 01:09:16 -05:00
khash Move 3rd party libraries to "vendored" directory 2026-03-02 01:09:16 -05:00
libfuse fix data race on _threads between constructor and autoscale workers 2026-04-10 22:24:14 -04:00
nonstd Update vendored/nonstd to latest and fix API usage 2026-03-02 01:09:16 -05:00
rapidhash Update vendored/rapidhash to v3 2026-03-02 01:09:16 -05:00
scope_guard Move scope_guard to vendored/ 2026-03-24 21:13:34 -04:00
subprocess Update vendored/subprocess to v2.5 2026-03-02 01:09:16 -05:00