- switched to asyncio-friendly test running and coverage gathering

  while prompting for tests, so that they don't block the event loop.
  This should avoid the OpenAI API timeouts we occasionally saw;

src/coverup/coverup.py

@@ -498,12 +498,14 @@ def pl(item, singular, plural = None):
                 return False # not finished: needs a missing module
 
         try:
-            result = measure_coverage(test=last_test, tests_dir=args.tests_dir, pytest_args=args.pytest_args,
-                                      log_write=lambda msg: log_write(seg, msg))
+            result = await measure_test_coverage(test=last_test, tests_dir=args.tests_dir, pytest_args=args.pytest_args,
+                                                 log_write=lambda msg: log_write(seg, msg))
 
         except subprocess.TimeoutExpired:
             log_write(seg, "measure_coverage timed out")
-            return False # try again next time
+            # FIXME is the built-in timeout reasonable? Do we prompt for a faster test?
+            # We don't want slow tests, but there may not be any way around it.
+            return True
 
         except subprocess.CalledProcessError as e:
             state.inc_counter('F')

src/coverup/testrunner.py

@@ -7,9 +7,10 @@
 import json
 import re
 from .delta import DeltaDebugger, _compact
+from .utils import subprocess_run
 
 
-def measure_coverage(*, test: str, tests_dir: Path, pytest_args='', log_write=None):
+async def measure_test_coverage(*, test: str, tests_dir: Path, pytest_args='', log_write=None):
     """Runs a given test and returns the coverage obtained."""
     with tempfile.NamedTemporaryFile(prefix="tmp_test_", suffix='.py',
                                      dir=str(tests_dir), mode="w") as t:
@@ -18,9 +19,9 @@ def measure_coverage(*, test: str, tests_dir: Path, pytest_args='', log_write=No
 
         with tempfile.NamedTemporaryFile() as j:
             # -qq to cut down on tokens
-            p = subprocess.run((f"{sys.executable} -m slipcover --branch --json --out {j.name} " +
-                                f"-m pytest {pytest_args} -qq --disable-warnings {t.name}").split(),
-                               check=True, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, timeout=60)
+            p = await subprocess_run((f"{sys.executable} -m slipcover --branch --json --out {j.name} " +
+                                      f"-m pytest {pytest_args} -qq --disable-warnings {t.name}").split(),
+                                      check=True, timeout=60)
             if log_write:
                 log_write(str(p.stdout, 'UTF-8'))