[3.12] pythongh-83434: Sync libregrtest and test_regrtest with the ma…

…in branch (python#117250)

* pythongh-115122: Add --bisect option to regrtest (python#115123)

* test.bisect_cmd now exit with code 0 on success, and code 1 on
  failure. Before, it was the opposite.
* test.bisect_cmd now runs the test worker process with
  -X faulthandler.
* regrtest RunTests: Add create_python_cmd() and bisect_cmd()
  methods.

(cherry picked from commit 1e5719a)

* pythongh-115720: Show number of leaks in huntrleaks progress reports (pythonGH-115726)

Instead of showing a dot for each iteration, show:
- '.' for zero (on negative) leaks
- number of leaks for 1-9
- 'X' if there are more leaks

This allows more rapid iteration: when bisecting, I don't need
to wait for the final report to see if the test still leaks.

Also, show the full result if there are any non-zero entries.
This shows negative entries, for the unfortunate cases where
a reference is created and cleaned up in different runs.

Test *failure* is still determined by the existing heuristic.

(cherry picked from commit af5f9d6)

* pythongh-83434: Disable XML in regrtest when -R option is used (python#117232)

(cherry picked from commit d52bdfb)

---------

Co-authored-by: Petr Viktorin <encukou@gmail.com>

Lib/test/bisect_cmd.py

@@ -51,6 +51,7 @@ def python_cmd():
     cmd = [sys.executable]
     cmd.extend(subprocess._args_from_interpreter_flags())
     cmd.extend(subprocess._optim_args_from_interpreter_flags())
+    cmd.extend(('-X', 'faulthandler'))
     return cmd
 
 
@@ -77,9 +78,13 @@ def run_tests(args, tests, huntrleaks=None):
         write_tests(tmp, tests)
 
         cmd = python_cmd()
-        cmd.extend(['-m', 'test', '--matchfile', tmp])
+        cmd.extend(['-u', '-m', 'test', '--matchfile', tmp])
         cmd.extend(args.test_args)
         print("+ %s" % format_shell_args(cmd))
+
+        sys.stdout.flush()
+        sys.stderr.flush()
+
         proc = subprocess.run(cmd)
         return proc.returncode
     finally:
@@ -137,8 +142,8 @@ def main():
             ntest = max(ntest // 2, 1)
             subtests = random.sample(tests, ntest)
 
-            print("[+] Iteration %s: run %s tests/%s"
-                  % (iteration, len(subtests), len(tests)))
+            print(f"[+] Iteration {iteration}/{args.max_iter}: "
+                  f"run {len(subtests)} tests/{len(tests)}")
             print()
 
             exitcode = run_tests(args, subtests)
@@ -170,10 +175,10 @@ def main():
     if len(tests) <= args.max_tests:
         print("Bisection completed in %s iterations and %s"
               % (iteration, datetime.timedelta(seconds=dt)))
-        sys.exit(1)
     else:
         print("Bisection failed after %s iterations and %s"
               % (iteration, datetime.timedelta(seconds=dt)))
+        sys.exit(1)
 
 
 if __name__ == "__main__":

Lib/test/libregrtest/cmdline.py

@@ -173,6 +173,7 @@ def __init__(self, **kwargs) -> None:
         self.fail_rerun = False
         self.tempdir = None
         self._add_python_opts = True
+        self.xmlpath = None
 
         super().__init__(**kwargs)
 
@@ -350,6 +351,8 @@ def _create_parser():
                        help='override the working directory for the test run')
     group.add_argument('--cleanup', action='store_true',
                        help='remove old test_python_* directories')
+    group.add_argument('--bisect', action='store_true',
+                       help='if some tests fail, run test.bisect_cmd on them')
     group.add_argument('--dont-add-python-opts', dest='_add_python_opts',
                        action='store_false',
                        help="internal option, don't use it")
@@ -497,17 +500,28 @@ def _parse_args(args, **kwargs):
         ns.randomize = True
     if ns.verbose:
         ns.header = True
+
     # When -jN option is used, a worker process does not use --verbose3
     # and so -R 3:3 -jN --verbose3 just works as expected: there is no false
     # alarm about memory leak.
     if ns.huntrleaks and ns.verbose3 and ns.use_mp is None:
-        ns.verbose3 = False
         # run_single_test() replaces sys.stdout with io.StringIO if verbose3
         # is true. In this case, huntrleaks sees an write into StringIO as
         # a memory leak, whereas it is not (gh-71290).
+        ns.verbose3 = False
         print("WARNING: Disable --verbose3 because it's incompatible with "
               "--huntrleaks without -jN option",
               file=sys.stderr)
+
+    if ns.huntrleaks and ns.xmlpath:
+        # The XML data is written into a file outside runtest_refleak(), so
+        # it looks like a leak but it's not. Simply disable XML output when
+        # hunting for reference leaks (gh-83434).
+        ns.xmlpath = None
+        print("WARNING: Disable --junit-xml because it's incompatible "
+              "with --huntrleaks",
+              file=sys.stderr)
+
     if ns.forever:
         # --forever implies --failfast
         ns.failfast = True

Lib/test/libregrtest/main.py

@@ -6,8 +6,7 @@
 import sysconfig
 import time
 
-from test import support
-from test.support import os_helper, MS_WINDOWS
+from test.support import os_helper, MS_WINDOWS, flush_std_streams
 
 from .cmdline import _parse_args, Namespace
 from .findtests import findtests, split_test_packages, list_cases
@@ -74,6 +73,7 @@ def __init__(self, ns: Namespace, _add_python_opts: bool = False):
         self.want_cleanup: bool = ns.cleanup
         self.want_rerun: bool = ns.rerun
         self.want_run_leaks: bool = ns.runleaks
+        self.want_bisect: bool = ns.bisect
 
         self.ci_mode: bool = (ns.fast_ci or ns.slow_ci)
         self.want_add_python_opts: bool = (_add_python_opts
@@ -277,6 +277,55 @@ def rerun_failed_tests(self, runtests: RunTests):
 
         self.display_result(rerun_runtests)
 
+    def _run_bisect(self, runtests: RunTests, test: str, progress: str) -> bool:
+        print()
+        title = f"Bisect {test}"
+        if progress:
+            title = f"{title} ({progress})"
+        print(title)
+        print("#" * len(title))
+        print()
+
+        cmd = runtests.create_python_cmd()
+        cmd.extend([
+            "-u", "-m", "test.bisect_cmd",
+            # Limit to 25 iterations (instead of 100) to not abuse CI resources
+            "--max-iter", "25",
+            "-v",
+            # runtests.match_tests is not used (yet) for bisect_cmd -i arg
+        ])
+        cmd.extend(runtests.bisect_cmd_args())
+        cmd.append(test)
+        print("+", shlex.join(cmd), flush=True)
+
+        flush_std_streams()
+
+        import subprocess
+        proc = subprocess.run(cmd, timeout=runtests.timeout)
+        exitcode = proc.returncode
+
+        title = f"{title}: exit code {exitcode}"
+        print(title)
+        print("#" * len(title))
+        print(flush=True)
+
+        if exitcode:
+            print(f"Bisect failed with exit code {exitcode}")
+            return False
+
+        return True
+
+    def run_bisect(self, runtests: RunTests) -> None:
+        tests, _ = self.results.prepare_rerun(clear=False)
+
+        for index, name in enumerate(tests, 1):
+            if len(tests) > 1:
+                progress = f"{index}/{len(tests)}"
+            else:
+                progress = ""
+            if not self._run_bisect(runtests, name, progress):
+                return
+
     def display_result(self, runtests):
         # If running the test suite for PGO then no one cares about results.
         if runtests.pgo:
@@ -458,7 +507,7 @@ def _run_tests(self, selected: TestTuple, tests: TestList | None) -> int:
 
         setup_process()
 
-        if self.hunt_refleak and not self.num_workers:
+        if (runtests.hunt_refleak is not None) and (not self.num_workers):
             # gh-109739: WindowsLoadTracker thread interfers with refleak check
             use_load_tracker = False
         else:
@@ -478,6 +527,9 @@ def _run_tests(self, selected: TestTuple, tests: TestList | None) -> int:
 
             if self.want_rerun and self.results.need_rerun():
                 self.rerun_failed_tests(runtests)
+
+            if self.want_bisect and self.results.need_rerun():
+                self.run_bisect(runtests)
         finally:
             if use_load_tracker:
                 self.logger.stop_load_tracker()

Lib/test/libregrtest/refleak.py

@@ -87,9 +87,12 @@ def get_pooled_int(value):
     rc_before = alloc_before = fd_before = interned_before = 0
 
     if not quiet:
-        print("beginning", repcount, "repetitions", file=sys.stderr)
-        print(("1234567890"*(repcount//10 + 1))[:repcount], file=sys.stderr,
-              flush=True)
+        print("beginning", repcount, "repetitions. Showing number of leaks "
+                "(. for 0 or less, X for 10 or more)",
+              file=sys.stderr)
+        numbers = ("1234567890"*(repcount//10 + 1))[:repcount]
+        numbers = numbers[:warmups] + ':' + numbers[warmups:]
+        print(numbers, file=sys.stderr, flush=True)
 
     results = None
     dash_R_cleanup(fs, ps, pic, zdc, abcs)
@@ -110,13 +113,27 @@ def get_pooled_int(value):
         rc_after = gettotalrefcount() - interned_after * 2
         fd_after = fd_count()
 
-        if not quiet:
-            print('.', end='', file=sys.stderr, flush=True)
-
         rc_deltas[i] = get_pooled_int(rc_after - rc_before)
         alloc_deltas[i] = get_pooled_int(alloc_after - alloc_before)
         fd_deltas[i] = get_pooled_int(fd_after - fd_before)
 
+        if not quiet:
+            # use max, not sum, so total_leaks is one of the pooled ints
+            total_leaks = max(rc_deltas[i], alloc_deltas[i], fd_deltas[i])
+            if total_leaks <= 0:
+                symbol = '.'
+            elif total_leaks < 10:
+                symbol = (
+                    '.', '1', '2', '3', '4', '5', '6', '7', '8', '9',
+                    )[total_leaks]
+            else:
+                symbol = 'X'
+            if i == warmups:
+                print(' ', end='', file=sys.stderr, flush=True)
+            print(symbol, end='', file=sys.stderr, flush=True)
+            del total_leaks
+            del symbol
+
         alloc_before = alloc_after
         rc_before = rc_after
         fd_before = fd_after
@@ -152,14 +169,20 @@ def check_fd_deltas(deltas):
     ]:
         # ignore warmup runs
         deltas = deltas[warmups:]
-        if checker(deltas):
+        failing = checker(deltas)
+        suspicious = any(deltas)
+        if failing or suspicious:
             msg = '%s leaked %s %s, sum=%s' % (
                 test_name, deltas, item_name, sum(deltas))
-            print(msg, file=sys.stderr, flush=True)
-            with open(filename, "a", encoding="utf-8") as refrep:
-                print(msg, file=refrep)
-                refrep.flush()
-            failed = True
+            print(msg, end='', file=sys.stderr)
+            if failing:
+                print(file=sys.stderr, flush=True)
+                with open(filename, "a", encoding="utf-8") as refrep:
+                    print(msg, file=refrep)
+                    refrep.flush()
+                failed = True
+            else:
+                print(' (this is fine)', file=sys.stderr, flush=True)
     return (failed, results)
 
 

Lib/test/libregrtest/results.py

@@ -129,7 +129,7 @@ def accumulate_result(self, result: TestResult, runtests: RunTests):
     def need_rerun(self):
         return bool(self.rerun_results)
 
-    def prepare_rerun(self) -> tuple[TestTuple, FilterDict]:
+    def prepare_rerun(self, *, clear: bool = True) -> tuple[TestTuple, FilterDict]:
         tests: TestList = []
         match_tests_dict = {}
         for result in self.rerun_results:
@@ -140,11 +140,12 @@ def prepare_rerun(self) -> tuple[TestTuple, FilterDict]:
             if match_tests:
                 match_tests_dict[result.test_name] = match_tests
 
-        # Clear previously failed tests
-        self.rerun_bad.extend(self.bad)
-        self.bad.clear()
-        self.env_changed.clear()
-        self.rerun_results.clear()
+        if clear:
+            # Clear previously failed tests
+            self.rerun_bad.extend(self.bad)
+            self.bad.clear()
+            self.env_changed.clear()
+            self.rerun_results.clear()
 
         return (tuple(tests), match_tests_dict)
 

Lib/test/libregrtest/runtests.py

@@ -2,7 +2,9 @@
 import dataclasses
 import json
 import os
+import shlex
 import subprocess
+import sys
 from typing import Any
 
 from test import support
@@ -67,6 +69,11 @@ class HuntRefleak:
     runs: int
     filename: StrPath
 
+    def bisect_cmd_args(self) -> list[str]:
+        # Ignore filename since it can contain colon (":"),
+        # and usually it's not used. Use the default filename.
+        return ["-R", f"{self.warmups}:{self.runs}:"]
+
 
 @dataclasses.dataclass(slots=True, frozen=True)
 class RunTests:
@@ -136,6 +143,47 @@ def json_file_use_stdout(self) -> bool:
             or support.is_wasi
         )
 
+    def create_python_cmd(self) -> list[str]:
+        python_opts = support.args_from_interpreter_flags()
+        if self.python_cmd is not None:
+            executable = self.python_cmd
+            # Remove -E option, since --python=COMMAND can set PYTHON
+            # environment variables, such as PYTHONPATH, in the worker
+            # process.
+            python_opts = [opt for opt in python_opts if opt != "-E"]
+        else:
+            executable = (sys.executable,)
+        cmd = [*executable, *python_opts]
+        if '-u' not in python_opts:
+            cmd.append('-u')  # Unbuffered stdout and stderr
+        return cmd
+
+    def bisect_cmd_args(self) -> list[str]:
+        args = []
+        if self.fail_fast:
+            args.append("--failfast")
+        if self.fail_env_changed:
+            args.append("--fail-env-changed")
+        if self.timeout:
+            args.append(f"--timeout={self.timeout}")
+        if self.hunt_refleak is not None:
+            args.extend(self.hunt_refleak.bisect_cmd_args())
+        if self.test_dir:
+            args.extend(("--testdir", self.test_dir))
+        if self.memory_limit:
+            args.extend(("--memlimit", self.memory_limit))
+        if self.gc_threshold:
+            args.append(f"--threshold={self.gc_threshold}")
+        if self.use_resources:
+            args.extend(("-u", ','.join(self.use_resources)))
+        if self.python_cmd:
+            cmd = shlex.join(self.python_cmd)
+            args.extend(("--python", cmd))
+        if self.randomize:
+            args.append(f"--randomize")
+        args.append(f"--randseed={self.random_seed}")
+        return args
+
 
 @dataclasses.dataclass(slots=True, frozen=True)
 class WorkerRunTests(RunTests):

Lib/test/libregrtest/worker.py

@@ -3,7 +3,6 @@
 import os
 from typing import Any, NoReturn
 
-from test import support
 from test.support import os_helper
 
 from .setup import setup_process, setup_test_dir
@@ -19,21 +18,10 @@
 
 def create_worker_process(runtests: WorkerRunTests, output_fd: int,
                           tmp_dir: StrPath | None = None) -> subprocess.Popen:
-    python_cmd = runtests.python_cmd
     worker_json = runtests.as_json()
 
-    python_opts = support.args_from_interpreter_flags()
-    if python_cmd is not None:
-        executable = python_cmd
-        # Remove -E option, since --python=COMMAND can set PYTHON environment
-        # variables, such as PYTHONPATH, in the worker process.
-        python_opts = [opt for opt in python_opts if opt != "-E"]
-    else:
-        executable = (sys.executable,)
-    cmd = [*executable, *python_opts,
-           '-u',    # Unbuffered stdout and stderr
-           '-m', 'test.libregrtest.worker',
-           worker_json]
+    cmd = runtests.create_python_cmd()
+    cmd.extend(['-m', 'test.libregrtest.worker', worker_json])
 
     env = dict(os.environ)
     if tmp_dir is not None: