Remove deprecated unbounded queue

[A5rocks]

Very much based off of #937; fixes #2922; should be in the next 0.x.0 release so don't merge this yet!

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 100.00000%. Comparing base (a670d60) to head (68435cd).
Report is 13 commits behind head on main.

Additional details and impacted files

@@               Coverage Diff                @@
##                 main        #2925    +/-   ##
================================================
  Coverage   100.00000%   100.00000%            
================================================
  Files             124          122     -2     
  Lines           18427        18313   -114     
  Branches         1215         1206     -9     
================================================
- Hits            18427        18313   -114     

Files with missing lines	Coverage Δ
src/trio/_core/__init__.py	100.00000% <ø> (ø)
src/trio/_core/_io_kqueue.py	100.00000% <100.00000%> (ø)
src/trio/_core/_io_windows.py	100.00000% <100.00000%> (ø)
src/trio/_core/_parking_lot.py	100.00000% <ø> (ø)
src/trio/_core/_tests/test_io.py	100.00000% <100.00000%> (ø)
src/trio/_core/_tests/test_windows.py	100.00000% <100.00000%> (ø)
src/trio/_tools/gen_exports.py	100.00000% <ø> (ø)
src/trio/lowlevel.py	100.00000% <ø> (ø)

... and 1 file with indirect coverage changes

[CoolCat467]

The grammar here is a bit odd. I would change it to something like
"but in the future if we ever support task priorities or
fair scheduling it should be fairly simple."

[A5rocks]

Huh actually I'm just realizing this is potentially a compat break. I can work around this using a kwarg but I need to see if this currently warns or not...

[Zac-HD]

Private subclass which adds the unbounded queue methods but warns if you call them?

[graingert]

it's possible for this to fail if someone calls:

with trio.lowlevel.monitor_kevent() as q:
    q.close()

we should either catch the exception here or yield a clone

I'd add it to this PR for you but I just cannot work out how to cover this line - it should be easy but I'm new to kqueue so might be missing something

here's my attempt at a test - but it times out 009cbb5 (#2925)

[graingert]

something like this:

from 9ddb5ecc47dda3a7502851967bb67247a99aefe5 mon sep 17 00:00:00 2001
from: thomas grainger <tagrain@gmail.com>
date: tue, 24 dec 2024 08:24:41 +0000
subject: [patch] make sure closing monitor queues is a noop

---
 src/trio/_core/_io_kqueue.py  | 12 ++++++------
 src/trio/_core/_io_windows.py | 13 +++++++------
 2 files changed, 13 insertions(+), 12 deletions(-)

diff --git a/src/trio/_core/_io_kqueue.py b/src/trio/_core/_io_kqueue.py
index fed4da83..08b5d966 100644
--- a/src/trio/_core/_io_kqueue.py
+++ b/src/trio/_core/_io_kqueue.py
@@ -137,12 +137,12 @@ class kqueueiomanager:
                 "attempt to register multiple listeners for same ident/filter pair",
             )
         send, recv = open_memory_channel[select.kevent](math.inf)
-        self._registered[key] = send
-        try:
-            yield recv
-        finally:
-            send.close()
-            self._registered.pop(key, none)
+        with send, recv:
+            self._registered[key] = send
+            try:
+                yield recv.clone()
+            finally:
+                self._registered.pop(key, none)
 
     @_public
     async def wait_kevent(
diff --git a/src/trio/_core/_io_windows.py b/src/trio/_core/_io_windows.py
index 4676e6c5..824e91f5 100644
--- a/src/trio/_core/_io_windows.py
+++ b/src/trio/_core/_io_windows.py
@@ -1040,9 +1040,10 @@ class windowsiomanager:
 
         key = next(self._completion_key_counter)
         send, recv = open_memory_channel[object](math.inf)
-        self._completion_key_queues[key] = send
-        try:
-            yield (key, recv)
-        finally:
-            send.close()
-            del self._completion_key_queues[key]
+        with send, recv:
+            self._completion_key_queues[key] = send
+            try:
+                yield (key, recv.clone())
+            finally:
+                del self._completion_key_queues[key]
-- 
2.43.0

[A5rocks]

@graingert the reason this stalled (and the reason I think your latest changes don't quite work) is because we need to provide a layer that has both memory channel and unbounded queue apis. Alternatively, a boolean kwarg to switch over.

Otherwise, we're breaking backwards compatibility.

[graingert]

These apis were sketches barely documented and not suitable for production use, they also warned if they were used. I think it's acceptable to break them in this case.