From 2efa0b6b439cde778d8af4a611c1070860b84d29 Mon Sep 17 00:00:00 2001
From: Adrian Chaves <adrian@zyte.com>
Date: Fri, 31 Jul 2026 15:25:08 +0200
Subject: [PATCH] Improve test coverage

---
 scrapy/pqueues.py     | 53 ++++++++++++++++++-------------------------
 tests/test_pqueues.py | 34 +++++++++++++++++++++++++++
 2 files changed, 56 insertions(+), 31 deletions(-)

diff --git a/scrapy/pqueues.py b/scrapy/pqueues.py
index abbcbb6da..3bdc7857d 100644
--- a/scrapy/pqueues.py
+++ b/scrapy/pqueues.py
@@ -186,30 +186,27 @@ class ScrapyPriorityQueue:
                 self.queues[priority] = self.qfactory(priority)
             q = self.queues[priority]
         q.push(request)  # this may fail (eg. serialization error)
-        if self.curprio is None or priority < self.curprio:
+        # A queue class may drop a request instead of storing it, and a
+        # priority with an empty queue must not become the current one, since
+        # pop() and peek() expect to find a request there.
+        if q and (self.curprio is None or priority < self.curprio):
             self.curprio = priority
 
+    def _curqueue(self) -> QueueProtocol:
+        assert self.curprio is not None
+        # Whichever dict holds a non-empty queue at the current priority. The
+        # other one may hold an empty queue left behind by a failed push, e.g.
+        # on a serialization error, which would hide the next request.
+        return self.queues.get(self.curprio) or self._start_queues[self.curprio]
+
     def pop(self) -> Request | None:
-        while self.curprio is not None:
-            for queues in (self.queues, self._start_queues):
-                q = queues.get(self.curprio)
-                # An empty queue can linger at a priority when a push failed
-                # after creating it, e.g. on a serialization error. Popping
-                # from it would return None and hide the request that the other
-                # dict may hold at the same priority.
-                if not q:
-                    continue
-                m = q.pop()
-                if not q:
-                    # The other dict may have no queue at this priority either,
-                    # and a curprio that neither dict has would make peek() come
-                    # up empty.
-                    self._update_curprio()
-                return m
-            # Nothing to pop at this priority: refreshing drops the empty
-            # leftovers and moves on to the next priority.
+        if self.curprio is None:
+            return None
+        q = self._curqueue()
+        request = q.pop()
+        if not q:
             self._update_curprio()
-        return None
+        return request
 
     def _update_curprio(self) -> None:
         # Keeping an empty queue would hold its storage open for nothing, and
@@ -233,17 +230,11 @@ class ScrapyPriorityQueue:
         """
         if self.curprio is None:
             return None
-        # The dicts are walked in the same order as in pop(), which is what
-        # makes the returned request the one that pop() then returns.
-        for queues in (self.queues, self._start_queues):
-            queue = queues.get(self.curprio)
-            # Empty queues can linger at a priority (see pop()), where they
-            # would hide the request that the other dict may hold at the same
-            # priority.
-            if queue:
-                # Protocols can't declare optional members
-                return cast("Request", queue.peek())  # type: ignore[attr-defined]
-        return None
+        # Picking the queue the same way as pop() is what makes the returned
+        # request the one that pop() then returns.
+        queue = self._curqueue()
+        # Protocols can't declare optional members
+        return cast("Request", queue.peek())  # type: ignore[attr-defined]
 
     def close(self) -> list[int]:
         active: set[int] = set()
diff --git a/tests/test_pqueues.py b/tests/test_pqueues.py
index 8ad10e1c6..a30f8f3b0 100644
--- a/tests/test_pqueues.py
+++ b/tests/test_pqueues.py
@@ -14,6 +14,15 @@ from scrapy.utils.test import get_crawler
 from tests.utils.downloader import MockDownloader
 
 
+class DroppingFifoMemoryQueue(FifoMemoryQueue):  # type: ignore[valid-type,misc]
+    """Queue that drops requests marked with ``drop`` in their metadata, the
+    way a custom queue class could drop requests it does not want to store."""
+
+    def push(self, request):
+        if not request.meta.get("drop"):
+            super().push(request)
+
+
 class TestPriorityQueue:
     def setup_method(self):
         self.crawler = get_crawler(Spider)
@@ -183,6 +192,31 @@ class TestPriorityQueue:
         assert queue.pop() is None
         assert not queue.close()
 
+    def test_push_into_a_queue_that_drops_the_request(self):
+        """A queue class may drop a request instead of storing it, and that
+        priority must not become the current one, or ``peek()`` and ``pop()``
+        would come up empty with requests still queued."""
+        if not hasattr(queuelib.queue.FifoMemoryQueue, "peek"):
+            pytest.skip("queuelib.queue.FifoMemoryQueue.peek is undefined")
+        temp_dir = tempfile.mkdtemp()
+        queue = ScrapyPriorityQueue.from_crawler(
+            self.crawler, DroppingFifoMemoryQueue, temp_dir
+        )
+        # A higher priority than the request below, so that it would become the
+        # current priority if it were stored.
+        queue.push(
+            Request("https://example.org/dropped", priority=1, meta={"drop": True})
+        )
+        kept = Request("https://example.org/kept")
+        queue.push(kept)
+
+        assert len(queue) == 1
+        assert queue.peek().url == kept.url
+        assert queue.pop().url == kept.url
+        assert queue.peek() is None
+        assert queue.pop() is None
+        assert not queue.close()
+
     def test_init_prios_without_a_restorable_queue(self):
         """A priority recorded on close may have nothing to restore, e.g. if
         its only request failed to serialize. ``curprio`` must not point at it,