refer to wiki instead of tag

Author: jakkdl

docs/source/reference-io.rst

@@ -556,7 +556,7 @@ going to be rough orders of magnitude to give you a sense of scale; if
 you need precise numbers for your environment, measure!)
 
 .. file.read benchmark is
-   https://github.com/python-trio/trio/blob/v0.27.0/notes-to-self/file-read-latency.py
+   https://github.com/python-trio/trio/wiki/notes-to-self#file-read-latencypy
 .. Numbers for spinning disks and SSDs are from taking a few random
    recent reviews from http://www.storagereview.com/best_drives and
    looking at their "4K Write Latency" test results for "Average MS"

src/trio/_core/_io_windows.py

@@ -152,7 +152,7 @@
 #
 # Unfortunately, the Windows kernel seems to have bugs if you try to issue
 # multiple simultaneous IOCTL_AFD_POLL operations on the same socket (see
-# https://github.com/python-trio/trio/blob/v0.27.0/notes-to-self/afd-lab.py).
+# https://github.com/python-trio/trio/wiki/notes-to-self#afd-labpy).
 # So if a user calls wait_readable and
 # wait_writable at the same time, we have to combine those into a single
 # IOCTL_AFD_POLL. This means we can't just use the wait_overlapped machinery.

src/trio/_highlevel_open_tcp_listeners.py

@@ -42,7 +42,7 @@
 # backlog just causes it to be silently truncated to the configured maximum,
 # so this is unnecessary -- we can just pass in "infinity" and get the maximum
 # that way. (Verified on Windows, Linux, macOS using
-# https://github.com/python-trio/trio/blob/v0.27.0/notes-to-self/measure-listen-backlog.py
+# https://github.com/python-trio/trio/wiki/notes-to-self#measure-listen-backlogpy
 def _compute_backlog(backlog: int | None) -> int:
     # Many systems (Linux, BSDs, ...) store the backlog in a uint16 and are
     # missing overflow protection, so we apply our own overflow protection.