Skip to content

fix(thread_pool): graceful-shutdown a retiring sync worker so reload() completes (true-async/server#93) #177

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
EdmondDantes merged 1 commit into from
Jul 4, 2026
Merged

fix(thread_pool): graceful-shutdown a retiring sync worker so reload() completes (true-async/server#93) #177

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
26 changes: 26 additions & 0 deletions thread_pool.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -164,6 +164,17 @@ typedef struct {
async_thread_channel_t *channel;
} thread_pool_worker_ctx_t;

/* GCC -O2 -Wmaybe-uninitialized misfires across this function's deeply nested
* zend_try blocks — flagging both loop-local task pointers and the zend_try
* macro's own __orig_bailout (all set before their SETJMP and never changed,
* hence defined per C 7.13.2.1). The macro lives in Zend/zend.h and cannot be
* annotated here, so suppress the false positive for this function only.
* GCC-only: Clang doesn't have this warning group and MSVC would warn on the
* unknown pragma (C4068). */
#if defined(__GNUC__) && !defined(__clang__)
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wmaybe-uninitialized"
#endif
static void thread_pool_worker_handler(zend_async_thread_event_t *event, void *ctx)
{
thread_pool_worker_ctx_t *wc = (thread_pool_worker_ctx_t *) ctx;
Expand Down Expand Up @@ -620,6 +631,18 @@ static void thread_pool_worker_handler(zend_async_thread_event_t *event, void *c
pool_scope = NULL;
}

/* Backstop for a retiring SYNC-mode worker: its tasks ran inline (no
* per-task coroutine scope), so any coroutine still alive is a straggler
* the bootloader spawned into the main scope (e.g. a DB-pool healthcheck
* timer) that the server-side scope drain never reaches. Its live reactor
* handle would keep AFTER_MAIN — and this worker's reload exit-token —
* hung forever, so force graceful shutdown to cancel it. Coroutine-mode
* pools instead drain their in-flight task coroutines below (see 077),
* so they must not take this path. */
if (!pool->coroutine_mode) {
start_graceful_shutdown();
}

ZEND_ASYNC_RUN_SCHEDULER_AFTER_MAIN(false);

/* AFTER_MAIN drained all task coroutines (and ran their dispose,
Expand Down Expand Up @@ -676,6 +699,9 @@ static void thread_pool_worker_handler(zend_async_thread_event_t *event, void *c

pefree(wc, 1);
}
#if defined(__GNUC__) && !defined(__clang__)
#pragma GCC diagnostic pop
#endif

///////////////////////////////////////////////////////////
/// Coroutine-mode task: spawn + extended_dispose
Expand Down
Loading