Merge branch 'main' into fix-missed-opt

Author: simonzgx

.ci/compute_projects.py

@@ -57,6 +57,13 @@
     ".ci": {"llvm", "clang", "lld", "lldb"},
 }
 
+# This mapping describes runtimes that should be enabled for a specific project,
+# but not necessarily run for testing. The only case of this currently is lldb
+# which needs some runtimes enabled for tests.
+DEPENDENT_RUNTIMES_TO_BUILD = {"lldb": {"libcxx", "libcxxabi", "libunwind"}}
+
+# This mapping describes runtimes that should be tested when the key project is
+# touched.
 DEPENDENT_RUNTIMES_TO_TEST = {"clang": {"libcxx", "libcxxabi", "libunwind"}}
 
 EXCLUDE_LINUX = {
@@ -180,16 +187,20 @@ def _compute_project_check_targets(projects_to_test: Set[str]) -> Set[str]:
 def _compute_runtimes_to_test(projects_to_test: Set[str]) -> Set[str]:
     runtimes_to_test = set()
     for project_to_test in projects_to_test:
-        if project_to_test not in DEPENDENT_RUNTIMES_TO_TEST:
-            continue
-        runtimes_to_test.update(DEPENDENT_RUNTIMES_TO_TEST[project_to_test])
+        if project_to_test in DEPENDENT_RUNTIMES_TO_TEST:
+            runtimes_to_test.update(DEPENDENT_RUNTIMES_TO_TEST[project_to_test])
+        if project_to_test in DEPENDENT_RUNTIMES_TO_BUILD:
+            runtimes_to_test.update(DEPENDENT_RUNTIMES_TO_BUILD[project_to_test])
     return runtimes_to_test
 
 
-def _compute_runtime_check_targets(runtimes_to_test: Set[str]) -> Set[str]:
+def _compute_runtime_check_targets(projects_to_test: Set[str]) -> Set[str]:
     check_targets = set()
-    for runtime_to_test in runtimes_to_test:
-        check_targets.add(PROJECT_CHECK_TARGETS[runtime_to_test])
+    for project_to_test in projects_to_test:
+        if project_to_test not in DEPENDENT_RUNTIMES_TO_TEST:
+            continue
+        for runtime_to_test in DEPENDENT_RUNTIMES_TO_TEST[project_to_test]:
+            check_targets.add(PROJECT_CHECK_TARGETS[runtime_to_test])
     return check_targets
 
 
@@ -216,16 +227,16 @@ def get_env_variables(modified_files: list[str], platform: str) -> Set[str]:
     projects_to_test = _compute_projects_to_test(modified_projects, platform)
     projects_to_build = _compute_projects_to_build(projects_to_test)
     projects_check_targets = _compute_project_check_targets(projects_to_test)
-    runtimes_to_test = _compute_runtimes_to_test(projects_to_test)
-    runtimes_check_targets = _compute_runtime_check_targets(runtimes_to_test)
+    runtimes_to_build = _compute_runtimes_to_test(projects_to_test)
+    runtimes_check_targets = _compute_runtime_check_targets(projects_to_test)
     # We use a semicolon to separate the projects/runtimes as they get passed
     # to the CMake invocation and thus we need to use the CMake list separator
     # (;). We use spaces to separate the check targets as they end up getting
     # passed to ninja.
     return {
         "projects_to_build": ";".join(sorted(projects_to_build)),
         "project_check_targets": " ".join(sorted(projects_check_targets)),
-        "runtimes_to_build": ";".join(sorted(runtimes_to_test)),
+        "runtimes_to_build": ";".join(sorted(runtimes_to_build)),
         "runtimes_check_targets": " ".join(sorted(runtimes_check_targets)),
     }
 

.ci/compute_projects_test.py

@@ -205,6 +205,17 @@ def test_ci(self):
             "check-cxx check-cxxabi check-unwind",
         )
 
+    def test_lldb(self):
+        env_variables = compute_projects.get_env_variables(
+            ["lldb/CMakeLists.txt"], "Linux"
+        )
+        self.assertEqual(env_variables["projects_to_build"], "clang;lldb;llvm")
+        self.assertEqual(env_variables["project_check_targets"], "check-lldb")
+        self.assertEqual(
+            env_variables["runtimes_to_build"], "libcxx;libcxxabi;libunwind"
+        )
+        self.assertEqual(env_variables["runtimes_check_targets"], "")
+
 
 if __name__ == "__main__":
     unittest.main()

.ci/metrics/metrics.py

@@ -1,12 +1,9 @@
 import collections
 import datetime
-import dateutil
 import github
-import json
 import logging
 import os
 import requests
-import sys
 import time
 
 from dataclasses import dataclass
@@ -29,8 +26,8 @@
 # name.
 GITHUB_JOB_TO_TRACK = {
     "github_llvm_premerge_checks": {
-        "Build and Test Linux (Test Only - Please Ignore Results)": "premerge_linux",
-        "Build and Test Windows (Test Only - Please Ignore Results)": "premerge_windows",
+        "Build and Test Linux": "premerge_linux",
+        "Build and Test Windows": "premerge_windows",
     }
 }
 
@@ -55,18 +52,6 @@
 # by trial and error).
 GRAFANA_METRIC_MAX_AGE_MN = 120
 
-# Lists the BuildKite jobs we want to track. Maps the BuildKite job name to
-# the metric name in Grafana. This is important not to lose metrics history
-# if the workflow name changes.
-BUILDKITE_WORKFLOW_TO_TRACK = {
-    ":linux: Linux x64": "buildkite_linux",
-    ":windows: Windows x64": "buildkite_windows",
-}
-
-# Number of builds to fetch per page. Since we scrape regularly, this can
-# remain small.
-BUILDKITE_GRAPHQL_BUILDS_PER_PAGE = 50
-
 
 @dataclass
 class JobMetrics:
@@ -86,181 +71,6 @@ class GaugeMetric:
     time_ns: int
 
 
-def buildkite_fetch_page_build_list(
-    buildkite_token: str, after_cursor: str = None
-) -> list[dict[str, str]]:
-    """Fetches a page of the build list using the GraphQL BuildKite API.
-
-    Returns the BUILDKITE_GRAPHQL_BUILDS_PER_PAGE last running/queued builds,
-    or the BUILDKITE_GRAPHQL_BUILDS_PER_PAGE running/queued builds
-    older than the one pointer by |after_cursor| if provided.
-    The |after_cursor| value is taken from the previous page returned by the
-    API.
-
-    Args:
-      buildkite_token: the secret token to authenticate GraphQL requests.
-      after_cursor: cursor after which to start the page fetch.
-
-    Returns:
-      The most recent builds after cursor (if set) with the following format:
-      [
-        {
-            "cursor": <value>,
-            "number": <build-number>,
-        }
-      ]
-    """
-
-    BUILDKITE_GRAPHQL_QUERY = """
-    query OrganizationShowQuery {{
-      organization(slug: "llvm-project") {{
-        pipelines(search: "Github pull requests", first: 1) {{
-          edges {{
-            node {{
-              builds (state: [CANCELING, CREATING, FAILING, RUNNING], first: {PAGE_SIZE}, after: {AFTER}) {{
-                edges {{
-                  cursor
-                  node {{
-                    number
-                  }}
-                }}
-              }}
-            }}
-          }}
-        }}
-      }}
-    }}
-    """
-    query = BUILDKITE_GRAPHQL_QUERY.format(
-        PAGE_SIZE=BUILDKITE_GRAPHQL_BUILDS_PER_PAGE,
-        AFTER="null" if after_cursor is None else '"{}"'.format(after_cursor),
-    )
-    query = json.dumps({"query": query})
-    url = "https://graphql.buildkite.com/v1"
-    headers = {
-        "Authorization": "Bearer " + buildkite_token,
-        "Content-Type": "application/json",
-    }
-    data = requests.post(url, data=query, headers=headers).json()
-    # De-nest the build list.
-    if "errors" in data:
-        logging.info("Failed to fetch BuildKite jobs: {}".format(data["errors"]))
-        return []
-    builds = data["data"]["organization"]["pipelines"]["edges"][0]["node"]["builds"][
-        "edges"
-    ]
-    # Fold cursor info into the node dictionnary.
-    return [{**x["node"], "cursor": x["cursor"]} for x in builds]
-
-
-def buildkite_get_build_info(build_number: str) -> dict:
-    """Returns all the info associated with the provided build number.
-
-    Note: for unknown reasons, graphql returns no jobs for a given build,
-    while this endpoint does, hence why this uses this API instead of graphql.
-
-    Args:
-      build_number: which build number to fetch info for.
-
-    Returns:
-      The info for the target build, a JSON dictionnary.
-    """
-
-    URL = "https://buildkite.com/llvm-project/github-pull-requests/builds/{}.json"
-    return requests.get(URL.format(build_number)).json()
-
-
-def buildkite_get_incomplete_tasks(buildkite_token: str) -> list:
-    """Returns all the running/pending BuildKite builds.
-
-    Args:
-     buildkite_token: the secret token to authenticate GraphQL requests.
-     last_cursor: the cursor to stop at if set. If None, a full page is fetched.
-    """
-    output = []
-    cursor = None
-    while True:
-        page = buildkite_fetch_page_build_list(buildkite_token, cursor)
-        if len(page) == 0:
-            break
-        cursor = page[-1]["cursor"]
-        output += page
-    return output
-
-
-def buildkite_get_metrics(
-    buildkite_token: str, previously_incomplete: set[int]
-) -> (list[JobMetrics], set[int]):
-    """Returns a tuple with:
-
-    - the metrics recorded for newly completed workflow jobs.
-    - the set of workflow still running now.
-
-    Args:
-      buildkite_token: the secret token to authenticate GraphQL requests.
-        previously_incomplete: the set of running workflows the last time this
-        function was called.
-    """
-
-    running_builds = buildkite_get_incomplete_tasks(buildkite_token)
-    incomplete_now = set([x["number"] for x in running_builds])
-    output = []
-
-    for build_id in previously_incomplete:
-        if build_id in incomplete_now:
-            continue
-
-        info = buildkite_get_build_info(build_id)
-        metric_timestamp = dateutil.parser.isoparse(info["finished_at"])
-        for job in info["jobs"]:
-            # This workflow is not interesting to us.
-            if job["name"] not in BUILDKITE_WORKFLOW_TO_TRACK:
-                continue
-
-            # Don't count canceled jobs.
-            if job["canceled_at"]:
-                continue
-
-            created_at = dateutil.parser.isoparse(job["created_at"])
-            scheduled_at = dateutil.parser.isoparse(job["scheduled_at"])
-            started_at = dateutil.parser.isoparse(job["started_at"])
-            finished_at = dateutil.parser.isoparse(job["finished_at"])
-
-            job_name = BUILDKITE_WORKFLOW_TO_TRACK[job["name"]]
-            queue_time = (started_at - scheduled_at).seconds
-            run_time = (finished_at - started_at).seconds
-            status = bool(job["passed"])
-
-            # Grafana will refuse to ingest metrics older than ~2 hours, so we
-            # should avoid sending historical data.
-            metric_age_mn = (
-                datetime.datetime.now(datetime.timezone.utc) - metric_timestamp
-            ).total_seconds() / 60
-            if metric_age_mn > GRAFANA_METRIC_MAX_AGE_MN:
-                logging.warning(
-                    f"Job {job['name']} from workflow {build_id} dropped due"
-                    + f" to staleness: {metric_age_mn}mn old."
-                )
-                continue
-
-            metric_timestamp_ns = int(metric_timestamp.timestamp()) * 10**9
-            workflow_id = build_id
-            workflow_name = "Github pull requests"
-            output.append(
-                JobMetrics(
-                    job_name,
-                    queue_time,
-                    run_time,
-                    status,
-                    metric_timestamp_ns,
-                    workflow_id,
-                    workflow_name,
-                )
-            )
-
-    return output, incomplete_now
-
-
 def github_get_metrics(
     github_repo: github.Repository, last_workflows_seen_as_completed: set[int]
 ) -> tuple[list[JobMetrics], int]:
@@ -349,19 +159,7 @@ def github_get_metrics(
                     running_count[metric_name] += 1
                 continue
 
-            job_result = int(job.conclusion == "success")
-            if job_result:
-                # We still might want to mark the job as a failure if one of the steps
-                # failed. This is required due to use setting continue-on-error in
-                # the premerge pipeline to prevent sending emails while we are
-                # testing the infrastructure.
-                # TODO(boomanaiden154): Remove this once the premerge pipeline is no
-                # longer in a testing state and we can directly assert the workflow
-                # result.
-                for step in job.steps:
-                    if step.conclusion != "success" and step.conclusion != "skipped":
-                        job_result = 0
-                        break
+            job_result = int(job.conclusion == "success" or job.conclusion == "skipped")
 
             created_at = job.created_at
             started_at = job.started_at
@@ -490,17 +288,13 @@ def upload_metrics(workflow_metrics, metrics_userid, api_key):
 def main():
     # Authenticate with Github
     github_auth = Auth.Token(os.environ["GITHUB_TOKEN"])
-    buildkite_token = os.environ["BUILDKITE_TOKEN"]
     grafana_api_key = os.environ["GRAFANA_API_KEY"]
     grafana_metrics_userid = os.environ["GRAFANA_METRICS_USERID"]
 
     # The last workflow this script processed.
     # Because the Github queries are broken, we'll simply log a 'processed'
     # bit for the last COUNT_TO_PROCESS workflows.
     gh_last_workflows_seen_as_completed = set()
-    # Stores the list of pending/running builds in BuildKite we need to check
-    # at the next iteration.
-    bk_incomplete = set()
 
     # Enter the main loop. Every five minutes we wake up and dump metrics for
     # the relevant jobs.
@@ -512,13 +306,8 @@ def main():
             github_repo, gh_last_workflows_seen_as_completed
         )
 
-        bk_metrics, bk_incomplete = buildkite_get_metrics(
-            buildkite_token, bk_incomplete
-        )
-
-        metrics = gh_metrics + bk_metrics
-        upload_metrics(metrics, grafana_metrics_userid, grafana_api_key)
-        logging.info(f"Uploaded {len(metrics)} metrics")
+        upload_metrics(gh_metrics, grafana_metrics_userid, grafana_api_key)
+        logging.info(f"Uploaded {len(gh_metrics)} metrics")
 
         time.sleep(SCRAPE_INTERVAL_SECONDS)