[Utils] Add new --update-tests flag to llvm-lit (#108425)

This adds a flag to lit for detecting and updating failing tests when
possible to do so automatically. The flag uses a plugin architecture
where config files can add additional auto-updaters for the types of
tests in the test suite. When a test fails with `--update-tests` enabled
lit passes the test RUN invocation and output to each registered test
updater until one of them signals that it updated the test (or all test
updaters have been run). As such it is the responsibility of the test
updater to only update tests where it is reasonably certain that it will
actually fix the test, or come close to doing so.

Initially adds support for UpdateVerifyTests and UpdateTestChecks. The
flag is currently only implemented for lit's internal shell, so
`--update-tests` implies `LIT_USE_INTERNAL_SHELL=1`.

Builds on work in #97369
Fixes #81320

Author: hnrklssn

lit/LitConfig.py

@@ -38,6 +38,7 @@ def __init__(
         parallelism_groups={},
         per_test_coverage=False,
         gtest_sharding=True,
+        update_tests=False,
     ):
         # The name of the test runner.
         self.progname = progname
@@ -89,6 +90,8 @@ def __init__(
         self.parallelism_groups = parallelism_groups
         self.per_test_coverage = per_test_coverage
         self.gtest_sharding = bool(gtest_sharding)
+        self.update_tests = update_tests
+        self.test_updaters = []
 
     @property
     def maxIndividualTestTime(self):

lit/TestRunner.py

@@ -1190,6 +1190,18 @@ def executeScriptInternal(
                 str(result.timeoutReached),
             )
 
+        if litConfig.update_tests:
+            for test_updater in litConfig.test_updaters:
+                try:
+                    update_output = test_updater(result, test)
+                except Exception as e:
+                    out += f"Exception occurred in test updater: {e}"
+                    continue
+                if update_output:
+                    for line in update_output.splitlines():
+                        out += f"# {line}\n"
+                    break
+
     return out, err, exitCode, timeoutInfo
 
 

lit/cl_arguments.py

@@ -204,6 +204,12 @@ def parse_args():
         action="store_true",
         help="Exit with status zero even if some tests fail",
     )
+    execution_group.add_argument(
+        "--update-tests",
+        dest="update_tests",
+        action="store_true",
+        help="Try to update regression tests to reflect current behavior, if possible",
+    )
     execution_test_time_group = execution_group.add_mutually_exclusive_group()
     execution_test_time_group.add_argument(
         "--skip-test-time-recording",

lit/llvm/config.py

@@ -64,12 +64,17 @@ def __init__(self, lit_config, config):
             self.with_environment("_TAG_REDIR_ERR", "TXT")
             self.with_environment("_CEE_RUNOPTS", "FILETAG(AUTOCVT,AUTOTAG) POSIX(ON)")
 
+        if lit_config.update_tests:
+            self.use_lit_shell = True
+
         # Choose between lit's internal shell pipeline runner and a real shell.
         # If LIT_USE_INTERNAL_SHELL is in the environment, we use that as an
         # override.
         lit_shell_env = os.environ.get("LIT_USE_INTERNAL_SHELL")
         if lit_shell_env:
             self.use_lit_shell = lit.util.pythonize_bool(lit_shell_env)
+            if not self.use_lit_shell and lit_config.update_tests:
+                print("note: --update-tests is not supported when using external shell")
 
         if not self.use_lit_shell:
             features.add("shell")

lit/main.py

@@ -42,6 +42,7 @@ def main(builtin_params={}):
         config_prefix=opts.configPrefix,
         per_test_coverage=opts.per_test_coverage,
         gtest_sharding=opts.gtest_sharding,
+        update_tests=opts.update_tests,
     )
 
     discovered_tests = lit.discovery.find_tests_for_inputs(