🐛 fix closed async backlog accounting

docs/api/async-logger-dropped-count.md

@@ -36,6 +36,7 @@ Detailed rules explaining key parameters and behaviors
 - The counter increases when overflow policy discards records.
 - The counter can also increase when `close(clear=true)` or `shutdown(clear=true)` abandons queued records.
 - It can also increase when shutdown fallback logic converts remaining pending records into dropped records during a clear-close path.
+- Later log attempts against an already closed queue do not add to this counter by themselves.
 - This is a cumulative counter for the lifetime of the logger value.
 - Use this helper when you need a focused loss metric rather than a full `state()` snapshot.
 

docs/api/async-logger-log.md

@@ -47,6 +47,7 @@ Detailed rules explaining key parameters and behaviors
 - Context fields, patch logic, and filter logic are applied before enqueue.
 - If timestamping is enabled, `@env.now()` is captured before the record enters the queue.
 - Overflow behavior depends on the configured `AsyncOverflowPolicy`.
+- Closed-on-log behavior is runtime-dependent: compatibility runtimes short-circuit before enqueue work, while native-worker runtimes may still reach the queue operations and then treat a closed queue as a non-accepted write.
 
 ### How to Use
 
@@ -82,6 +83,8 @@ e.g.:
 
 - If the logger is closed or overflow policy rejects the record, enqueue may not proceed as a normal accepted write.
 
+- A closed queue does not count as a newly accepted pending record.
+
 ### Notes
 
 1. Use this API when the call site needs full control instead of a fixed severity helper.

docs/api/async-logger-pending-count.md

@@ -36,6 +36,7 @@ Detailed rules explaining key parameters and behaviors
 - The count increases when records are accepted into the queue.
 - The count decreases as the worker drains records.
 - The count is also reset to `0` when queued records are abandoned through clear-close paths such as `close(clear=true)`.
+- Log attempts against a closed queue do not create new pending backlog.
 - This is a point-in-time metric and may change immediately after it is read.
 - Use this helper when a single backlog number is enough and a full `state()` snapshot is unnecessary.
 

src-async/BitLoggerAsync_test.mbt

@@ -91,6 +91,29 @@ async test "shutdown clear closes without worker startup" {
   inspect(logger.dropped_count(), content="1")
 }
 
+async test "closed blocking logger does not add pending count on later log attempts" {
+  let logger = async_logger(
+    @bitlogger.callback_sink(fn(_) {
+
+    }),
+    config=AsyncLoggerConfig::new(
+      max_pending=2,
+      overflow=AsyncOverflowPolicy::Blocking,
+    ),
+    min_level=@bitlogger.Level::Info,
+    target="async.closed.blocking",
+  )
+
+  logger.info("one")
+  logger.close(clear=true)
+  inspect(logger.pending_count(), content="0")
+  inspect(logger.dropped_count(), content="1")
+
+  logger.info("late")
+  inspect(logger.pending_count(), content="0")
+  inspect(logger.dropped_count(), content="1")
+}
+
 test "async logger config stringify roundtrips stable fields" {
   let text = stringify_async_logger_config(
     AsyncLoggerConfig::new(

src-async/async_logger_shared.mbt

@@ -254,11 +254,16 @@ pub async fn[S] AsyncLogger::log(
   } else {
     match self.overflow {
       AsyncOverflowPolicy::Blocking => {
-        self.queue.put(rec) catch {
-          err if err is AsyncLoggerClosed => ()
+        let accepted = (async fn() -> Bool raise {
+          self.queue.put(rec)
+          true
+        })() catch {
+          err if err is AsyncLoggerClosed => false
           err => raise err
         }
-        self.pending_count.val += 1
+        if accepted {
+          self.pending_count.val += 1
+        }
       }
       AsyncOverflowPolicy::DropOldest | AsyncOverflowPolicy::DropNewest => {
         self.dropped_count.val += 1