Forward-merge release/26.08 into main

ci/test_python_sklearn_examples.sh

@@ -1,5 +1,5 @@
 #!/bin/bash
-# SPDX-FileCopyrightText: Copyright (c) 2025-2026, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2025-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
 # Support invoking test script outside the script directory
@@ -30,12 +30,16 @@ timeout -v --signal=SIGINT --kill-after=60s 60m ./python/cuml/cuml_accel_tests/u
     --example-timeout=300 \
     --junitxml="${SKLEARN_EXAMPLES_JUNITXML}"
 
-# Per-example timeouts and network failures are reported as xfails. The
-# examples tests still require a healthy majority of examples to pass so
-# widespread regressions are not missed.
-rapids-logger "scikit-learn examples: require >=90% pass rate"
+# Per-example timeouts and network failures are reported as xfails. Network
+# failures do not exercise cuml.accel and are excluded from the pass-rate
+# denominator. Timeouts and all other xfails remain non-passing outcomes. The
+# lower total pass-rate threshold includes network failures to catch widespread
+# network outages.
+rapids-logger "scikit-learn examples: require >=90% pass rate and >=80% total"
 ./python/cuml/cuml_accel_tests/upstream/summarize-results.py \
     --fail-below 90 \
+    --total-fail-below 80 \
+    --exclude-xfail-reason "Network error:" \
     "${SKLEARN_EXAMPLES_JUNITXML}"
 
 rapids-logger "Test script exiting with value: $EXITCODE"

python/cuml/cuml_accel_tests/upstream/README.md

@@ -155,10 +155,14 @@ Useful `summarize-results.py` options:
 
 - `-v, --verbose`: Display detailed failure information.
 - `-f, --fail-below VALUE`: Set a minimum pass-rate threshold from 0 to 100.
+- `--total-fail-below VALUE`: Set a minimum total pass-rate threshold,
+  including outcomes excluded by `--exclude-xfail-reason`.
 - `--format FORMAT`: Output `summary`, `xfail_list`, or `traceback`.
 - `--limit N`: Limit output to the first `N` entries.
 - `--test-id-prefix PREFIX`: Prefix added to test IDs in generated output.
 - `-k, --filter PATTERN`: Filter tests by ID substring, case-insensitively.
+- `--exclude-xfail-reason TEXT`: Exclude xfails whose JUnit reason contains
+  `TEXT` from the pass-rate denominator. May be specified multiple times.
 - `--config FILE`: Load summary defaults from a config file, such as
   `scikit-learn/test_config.yaml` for scikit-learn tests.
 

python/cuml/cuml_accel_tests/upstream/summarize-results.py

@@ -1,5 +1,5 @@
 #!/usr/bin/env python3
-# SPDX-FileCopyrightText: Copyright (c) 2025-2026, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2025-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
 """Summarize test results from a JUnit XML report file."""
@@ -71,6 +71,14 @@ def parse_args():
         type=float,
         help="Minimum pass rate threshold [0-100] (default: 0)",
     )
+    parser.add_argument(
+        "--total-fail-below",
+        type=float,
+        help=(
+            "Minimum total pass rate threshold [0-100], including outcomes "
+            "excluded by --exclude-xfail-reason (default: disabled)"
+        ),
+    )
     parser.add_argument(
         "--format",
         choices=["summary", "xfail_list", "traceback"],
@@ -94,6 +102,16 @@ def parse_args():
         dest="filter_pattern",
         help="Filter tests by ID pattern (substring match, case-insensitive)",
     )
+    parser.add_argument(
+        "--exclude-xfail-reason",
+        action="append",
+        default=[],
+        metavar="TEXT",
+        help=(
+            "Exclude xfails whose JUnit reason contains TEXT from the pass-rate "
+            "denominator. May be specified multiple times."
+        ),
+    )
     args = parser.parse_args()
 
     # Load config if provided
@@ -137,6 +155,12 @@ def matches_filter(test_id, pattern):
     return pattern.lower() in test_id.lower()
 
 
+def matches_xfail_reason(skipped_elem, patterns):
+    """Return whether an xfail reason matches any exclusion pattern."""
+    message = skipped_elem.get("message", "")
+    return any(pattern in message for pattern in patterns)
+
+
 def get_test_results(testsuite, prefix: str = ""):
     """Extract test results from testsuite.
 
@@ -289,6 +313,8 @@ def main():
     """Main entry point."""
     args = parse_args()
     validate_threshold(args.fail_below)
+    if args.total_fail_below is not None:
+        validate_threshold(args.total_fail_below)
 
     if not args.report_file.exists():
         print(f"Error: Report file not found: {args.report_file}")
@@ -317,6 +343,7 @@ def main():
     regular_errors = 0
     regular_skipped = 0
     xfailed = 0
+    excluded_xfailed = 0
     xpassed_strict = 0
     xpassed_non_strict = 0
     for testcase in testsuite.findall(".//testcase"):
@@ -345,6 +372,10 @@ def main():
         elif skipped_elem is not None:
             if skipped_elem.get("type") == "pytest.xfail":
                 xfailed += 1
+                if matches_xfail_reason(
+                    skipped_elem, args.exclude_xfail_reason
+                ):
+                    excluded_xfailed += 1
             else:
                 regular_skipped += 1
 
@@ -358,7 +389,13 @@ def main():
         - xpassed_strict
         - xpassed_non_strict
     )
-    pass_rate = (passed / total_tests * 100) if total_tests > 0 else 0
+    pass_rate_denominator = total_tests - excluded_xfailed
+    pass_rate = (
+        passed / pass_rate_denominator * 100
+        if pass_rate_denominator > 0
+        else 0
+    )
+    total_pass_rate = passed / total_tests * 100 if total_tests > 0 else 0
 
     if args.format == "traceback":
         output = format_traceback_output(
@@ -407,11 +444,13 @@ def main():
         ["Passed:", str(passed)],
         ["Failed:", str(regular_failures)],
         ["XFailed:", str(xfailed)],
+        ["Excluded XFailed:", str(excluded_xfailed)],
         ["XPassed (strict):", str(xpassed_strict)],
         ["XPassed (non-strict):", str(xpassed_non_strict)],
         ["Errors:", str(regular_errors)],
         ["Skipped:", str(regular_skipped)],
         ["Pass Rate:", f"{pass_rate:.2f}%"],
+        ["Total Pass Rate:", f"{total_pass_rate:.2f}%"],
         ["Total Time:", f"{time:.2f}s"],
     ]
     for row in format_table(rows, "  "):
@@ -469,11 +508,23 @@ def main():
                     print(f'  "{test_id}"')
                     count += 1
 
+    threshold_failed = False
     if pass_rate < args.fail_below:
         print(
             f"\nError: Pass rate {pass_rate:.2f}% is below threshold "
             f"{args.fail_below}%"
         )
+        threshold_failed = True
+    if (
+        args.total_fail_below is not None
+        and total_pass_rate < args.total_fail_below
+    ):
+        print(
+            f"\nError: Total pass rate {total_pass_rate:.2f}% is below "
+            f"threshold {args.total_fail_below}%"
+        )
+        threshold_failed = True
+    if threshold_failed:
         sys.exit(1)
 
     sys.exit(0)