Loosen import-time gate on Windows and unpin pytest (#4532)

Author: h-mayorquin

.github/scripts/build_job_summary.py

@@ -3,6 +3,7 @@
 The input file is the output of the following command:
 pytest -vv --durations=0 --durations-min=0.001 > report.txt
 """
+import re
 from pathlib import Path
 import pandas as pd
 import sys
@@ -20,9 +21,12 @@
 start_index = all_lines.index(start_line) + 1
 
 last_index = next(index for index, line in enumerate(all_lines[start_index:]) if "===" in line) + start_index
-#last_index = all_lines.index(last_line)
 
-timing_info = all_lines[start_index:last_index]
+# Pytest 8.4+ appends a blank line and a "(N durations < Xs hidden.)" footer
+# inside the slowest-durations block. Keep only true duration rows shaped like
+# "0.123s call test_x.py::test_name".
+duration_line_re = re.compile(r"^\d+\.\d+s\s+(call|setup|teardown)\s")
+timing_info = [line for line in all_lines[start_index:last_index] if duration_line_re.match(line)]
 timing_column = [float(line.split("s")[0].rstrip()) for line in timing_info]
 type = [line.split("s")[1].rstrip() for line in timing_info]
 short_name = [line.rpartition('::')[2] for line in timing_info]

.github/scripts/import_test.py

@@ -45,14 +45,14 @@
         time_taken_list.append(time_taken)
 
     for time in time_taken_list:
-        import_time_threshold = 3.0  # Most of the times is sub-second but there outliers
+        # TODO: lower this back toward 3.0 s once the Windows runner outliers are diagnosed.
+        import_time_threshold = 6.0
         if time >= import_time_threshold:
             exceptions.append(
                 f"Importing {import_statement} took: {time:.2f} s. Should be <: {import_time_threshold} s."
             )
             break
 
-
     if time_taken_list:
         avg_time = sum(time_taken_list) / len(time_taken_list)
         std_time = math.sqrt(sum((x - avg_time) ** 2 for x in time_taken_list) / len(time_taken_list))
@@ -65,8 +65,9 @@
                 f"Importing {import_statement} took: {avg_time:.2f} s in average. Should be <: {import_time_threshold} s."
             )
 
+# This is displayed to GITHUB_STEP_SUMMARY. Print it before raising so the
+# per-sample table is available even when the average threshold is exceeded.
+print(markdown_output)
+
 if exceptions:
     raise Exception("\n".join(exceptions))
-
-# This is displayed to GITHUB_STEP_SUMMARY
-print(markdown_output)

pyproject.toml

@@ -123,7 +123,7 @@ metrics = [
 
 test_core = [
     "pandas<3",
-    "pytest<8.4.0",
+    "pytest",
     "psutil",
 
     # for github test : probeinterface and neo from master
@@ -154,7 +154,7 @@ test_preprocessing = [
 
 
 test = [
-    "pytest<8.4.0",
+    "pytest",
     "pytest-cov",
     "psutil",