charisma-cli 0.1.6__tar.gz → 0.2.0__tar.gz

{charisma_cli-0.1.6 → charisma_cli-0.2.0}/.gitignore

@@ -132,3 +132,4 @@ ltm/snapshots/*
 backend/coverage.json
 */__pycache__/*
 allure-results/*
+allure-results-live/

{charisma_cli-0.1.6 → charisma_cli-0.2.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: charisma-cli
-Version: 0.1.6
+Version: 0.2.0
 Summary: CLI tool that watches allure-results and streams test results + attachments to Charisma.
 Author: Charisma Team
 License-Expression: MIT

{charisma_cli-0.1.6 → charisma_cli-0.2.0}/pyproject.toml

@@ -5,7 +5,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "charisma-cli"
-version = "0.1.6"
+version = "0.2.0"
 description = "CLI tool that watches allure-results and streams test results + attachments to Charisma."
 readme = "README.md"
 license = "MIT"

{charisma_cli-0.1.6 → charisma_cli-0.2.0}/src/charisma_cli/__init__.py

@@ -1,3 +1,3 @@
 """charisma-cli: Stream allure results to Charisma in real time."""
 
-__version__ = "0.1.2"
+__version__ = "0.1.6"

{charisma_cli-0.1.6 → charisma_cli-0.2.0}/src/charisma_cli/main.py

@@ -16,6 +16,15 @@ from charisma_cli.subprocess_mgr import SubprocessManager
 from charisma_cli.uploader import Uploader
 from charisma_cli.watcher import ResultsWatcher, classify_file
 
+# Post-subprocess settle window for the watch command. allure-pytest under heavy
+# xdist can finish flushing its result files to disk seconds after the test
+# subprocess exits; settle_scan keeps polling until the directory is quiet for
+# SETTLE_QUIET_PERIOD or SETTLE_MAX_WAIT elapses. The window is deliberately
+# wider than settle_scan's own defaults (1s/15s) because a real CI burst can
+# land well past 15s — a too-short window is the watch results_sent=0 bug.
+SETTLE_QUIET_PERIOD = 2.0
+SETTLE_MAX_WAIT = 60.0
+
 
 @click.group()
 @click.version_option(version=__version__, prog_name="charismactl")
@@ -185,17 +194,19 @@ def watch(
     mgr.spawn(command)
     exit_code = mgr.wait()
 
-    # After subprocess exits (crash or normal): flush debouncing files, then
-    # keep scanning until the results directory settles. allure-pytest writes
-    # its result files in a burst during test-process teardown; under high
-    # xdist parallelism that burst can land AFTER the subprocess exits, so a
-    # single scan taken here races the OS flush and captures nothing (the
-    # watch results_sent=0 bug). settle_scan reconciles until the directory is
-    # quiet, capturing the whole burst before we drain.
+    # Reconcile the results directory after the subprocess exits. allure-pytest
+    # writes its result files in a burst during test-process teardown; under
+    # heavy xdist parallelism that burst can finish flushing to disk seconds
+    # after the subprocess returns. flush_pending drains any files still in the
+    # 500ms debounce window; settle_scan then polls the directory until it is
+    # quiet, capturing the whole late burst before we drain. Without a wide
+    # enough window the queue is drained empty and nothing is sent (the watch
+    # results_sent=0 bug). The shared seen-set keeps every file exactly-once.
     watcher.flush_pending()
-    watcher.settle_scan()
+    watcher.settle_scan(quiet_period=SETTLE_QUIET_PERIOD, max_wait=SETTLE_MAX_WAIT)
 
-    # Drain and close — stop watcher AFTER drain so all files are processed
+    # Drain and close — stop the watcher AFTER drain so the consumer's final
+    # queue flush still sees anything the observer enqueued during drain.
     uploader.drain(timeout=config.drain_timeout)
     watcher.stop()
     uploader.stop()

{charisma_cli-0.1.6 → charisma_cli-0.2.0}/src/charisma_cli/uploader.py

@@ -26,6 +26,11 @@ _BATCH_SIZE = 50
 _BATCH_TIMEOUT_SECONDS = 2.0
 _CONTAINER_ASSOCIATION_TIMEOUT = 60.0
 
+# Sentinel expectedTests for CLI launches: the total count is unknown at open
+# time (directory-watch streaming), so we send the backend maximum. This keeps
+# the launch in 'receiving' across all appends; close_launch() finalizes it.
+_UNKNOWN_EXPECTED_TESTS = 1_000_000
+
 
 def _epoch_ms_to_iso(epoch_ms: int) -> str:
     """Convert epoch milliseconds to ISO 8601 UTC string."""
@@ -173,9 +178,14 @@ class Uploader:
         """
         self.ensure_client()
 
+        # The CLI streams from a watched directory and does not know the total
+        # test count at open time. Send a high sentinel so the launch stays
+        # 'receiving' through every append; finalization is driven by
+        # close_launch() at the end of the run (or the stale-launch finalizer as
+        # a backstop), never by the append counter reaching expectedTests.
         payload: dict[str, Any] = {
             "projectAlias": self._config.project,
-            "expectedTests": 1000,
+            "expectedTests": _UNKNOWN_EXPECTED_TESTS,
         }
         if self._config.build_id:
             payload["buildId"] = self._config.build_id
@@ -306,16 +316,16 @@ class Uploader:
         if result.uuid:
             self._known_result_uuids.add(result.uuid)
 
-        # Build API payload
+        # Build API payload (snake_case wire format — shared TestResultInput contract)
         payload: dict[str, Any] = {
-            "testId": result.testId,
+            "test_id": result.testId,
             "outcome": result.outcome,
             "duration_ms": result.duration_ms,
         }
         if result.name:
             payload["name"] = result.name
         if result.full_name:
-            payload["fullName"] = result.full_name
+            payload["full_name"] = result.full_name
         if result.error_message:
             payload["error_message"] = result.error_message
         if result.stack_trace:
@@ -344,10 +354,11 @@ class Uploader:
         if result.ended_at is not None:
             payload["ended_at"] = _epoch_ms_to_iso(result.ended_at)
 
-        # Serialize Allure steps and links into labels dict (per design field mapping)
+        # Serialize Allure steps into the first-class `steps` array (StepInput
+        # wire shape). Previously packed as a JSON string in labels["steps"] —
+        # dropped in favor of the shared TestResultInput contract.
         steps = raw_data.get("steps")
         if steps:
-            # Normalize Allure step format to match the schema _parse_steps expects
             normalized_steps = []
             for step in steps:
                 status_details = step.get("statusDetails") or {}
@@ -359,17 +370,16 @@ class Uploader:
                     "duration_ms": (stop - start) if (start and stop) else None,
                     "error_message": status_details.get("message"),
                 })
-            labels_d = payload.get("labels", {})
-            labels_d["steps"] = json.dumps(normalized_steps)
-            payload["labels"] = labels_d
+            payload["steps"] = normalized_steps
 
+        # Links and description are label-like metadata not modeled by StepInput;
+        # keep them in labels (values are strings, matching labels: dict[str, str]).
         links = raw_data.get("links")
         if links:
             labels_d = payload.get("labels", {})
             labels_d["links"] = json.dumps(links)
             payload["labels"] = labels_d
 
-        # Add description to labels if present
         description = raw_data.get("description") or raw_data.get("descriptionHtml")
         if description:
             labels_d = payload.get("labels", {})

{charisma_cli-0.1.6 → charisma_cli-0.2.0}/src/charisma_cli/watcher.py

@@ -1,5 +1,6 @@
 """Watchdog observer, stability debounce, and file classification for allure-results."""
 
+import logging
 import os
 import time
 from pathlib import Path
@@ -12,6 +13,8 @@ from watchdog.observers import Observer
 from charisma_cli.config import Config
 from charisma_cli.models import FileCategory, FileEvent
 
+logger = logging.getLogger(__name__)
+
 _TWO_MB = 2 * 1024 * 1024
 _STABILITY_SECONDS = 0.5
 
@@ -306,8 +309,10 @@ class ResultsWatcher:
             Total number of new files enqueued across all iterations.
         """
         total = 0
-        deadline = time.monotonic() + max_wait
-        last_new_at = time.monotonic()
+        started_at = time.monotonic()
+        deadline = started_at + max_wait
+        last_new_at = started_at
+        settled = False
 
         while time.monotonic() < deadline:
             new_count = self.final_scan()
@@ -317,7 +322,27 @@ class ResultsWatcher:
                 last_new_at = now
             elif now - last_new_at >= quiet_period:
                 # No new files for a full quiet period — directory has settled.
+                settled = True
                 break
             time.sleep(poll_interval)
 
+        elapsed = time.monotonic() - started_at
+        if settled:
+            logger.info(
+                "Results directory settled after %.1fs (%d file(s) captured during settle)",
+                elapsed,
+                total,
+            )
+        else:
+            # Exited on max_wait, not quiescence: the directory was still
+            # producing files when we gave up. Surface this so an operator can
+            # see teardown burned the full window rather than settling — a
+            # signal that max_wait may be too short for this workload.
+            logger.warning(
+                "settle_scan hit max_wait (%.0fs) before the results directory "
+                "settled; captured %d file(s). Some late results may be unsent.",
+                max_wait,
+                total,
+            )
+
         return total

{charisma_cli-0.1.6 → charisma_cli-0.2.0}/tests/test_integration.py

@@ -2,6 +2,7 @@
 
 import json
 import sys
+import threading
 from pathlib import Path
 
 import httpx
@@ -228,10 +229,10 @@ class TestReportingFixesEndToEnd:
         # The skipped result was sent (not dropped)
         assert results_route.called
         sent_results = json.loads(results_route.calls[0].request.content)["results"]
-        outcomes = {r["testId"]: r["outcome"] for r in sent_results}
+        outcomes = {r["test_id"]: r["outcome"] for r in sent_results}
         assert outcomes.get("hist_skip") == "skipped"
         # duration defaulted to 0 for the missing stop
-        skip_payload = next(r for r in sent_results if r["testId"] == "hist_skip")
+        skip_payload = next(r for r in sent_results if r["test_id"] == "hist_skip")
         assert skip_payload["duration_ms"] == 0
 
         # environment.properties sent as variables on close
@@ -242,3 +243,132 @@ class TestReportingFixesEndToEnd:
             "aws_region": "us-west-2",
             "workers_number": "8",
         }
+
+
+class TestSettleScanCapturesLateTeardownBurst:
+    """Regression for the `watch results_sent=0` bug — isolates the true fix.
+
+    The one behavioral property that separates the old teardown from the new is
+    settle_scan's max_wait window. allure-pytest under heavy xdist can flush its
+    result files to disk SEVERAL SECONDS after the subprocess exits. The old
+    teardown used settle_scan()'s default max_wait=15.0; the new watch teardown
+    widens it to 60.0. A file that first lands on disk AFTER the window closes is
+    never enqueued and never sent — that is the production failure.
+
+    This test drives ResultsWatcher directly (not the full CLI) with scaled-down
+    timings so it is fast and deterministic: a result file is written by a
+    background timer at t=0.3s after settle_scan starts. With a SHORT window
+    (mimicking the old code) settle_scan gives up before the file lands and
+    enqueues nothing. With a LONGER window (mimicking the new code) it is still
+    polling when the file lands and enqueues it. No reliance on OS event buffers
+    or thread scheduling — only on-disk appearance time vs the window.
+    """
+
+    def _write_result_after(self, results_dir: Path, delay: float) -> threading.Timer:
+        """Schedule a valid Allure result file to appear after `delay` seconds."""
+
+        def _write() -> None:
+            (results_dir / "late-result.json").write_text(json.dumps({
+                "uuid": "late-1",
+                "historyId": "hist_late",
+                "fullName": "tests.test_late.test_case",
+                "name": "test_case",
+                "status": "passed",
+                "start": 1000,
+                "stop": 2000,
+                "labels": [],
+                "parameters": [],
+                "attachments": [],
+            }))
+
+        timer = threading.Timer(delay, _write)
+        timer.daemon = True
+        timer.start()
+        return timer
+
+    def test_short_window_misses_late_file_long_window_captures_it(
+        self, tmp_path: Path, default_config
+    ) -> None:
+        """settle_scan with a window shorter than the file's appearance misses it;
+        a window longer than the appearance captures it. This is exactly the
+        old (max_wait=15) vs new (max_wait=60) separation, scaled to sub-second."""
+        from queue import PriorityQueue
+
+        from charisma_cli.models import FileEvent
+        from charisma_cli.watcher import ResultsWatcher
+
+        # --- OLD behavior: window closes BEFORE the file lands (t=0.3s) ---
+        old_dir = tmp_path / "old"
+        old_dir.mkdir()
+        old_queue: PriorityQueue[FileEvent] = PriorityQueue()
+        old_watcher = ResultsWatcher(str(old_dir), old_queue, default_config)
+        t1 = self._write_result_after(old_dir, delay=0.3)
+        # Do not start the observer — isolate the scan behavior from OS events.
+        old_captured = old_watcher.settle_scan(quiet_period=0.1, max_wait=0.2)
+        t1.join()
+        assert old_captured == 0, "short window should give up before the late file lands"
+        assert old_queue.empty()
+
+        # --- NEW behavior: window stays open PAST the file landing (t=0.3s) ---
+        new_dir = tmp_path / "new"
+        new_dir.mkdir()
+        new_queue: PriorityQueue[FileEvent] = PriorityQueue()
+        new_watcher = ResultsWatcher(str(new_dir), new_queue, default_config)
+        t2 = self._write_result_after(new_dir, delay=0.3)
+        new_captured = new_watcher.settle_scan(quiet_period=0.3, max_wait=2.0)
+        t2.join()
+        assert new_captured == 1, "long window should still be polling when the late file lands"
+        assert new_queue.qsize() == 1
+
+
+class TestWatchStreamsPreexistingResult:
+    """Sanity E2E: watch streams a result already on disk through the launch API.
+
+    This does NOT isolate the old/new fix (any on-disk file is caught by both).
+    It guards the happy path: watch opens a launch, sends the result, closes,
+    and reports results_sent in the summary.
+    """
+
+    @respx.mock
+    def test_watch_sends_result_and_reports_summary(self, tmp_path: Path, monkeypatch) -> None:
+        results_dir = tmp_path / "allure-results"
+        results_dir.mkdir()
+        (results_dir / "pre-result.json").write_text(json.dumps({
+            "uuid": "pre-1",
+            "historyId": "hist_pre",
+            "fullName": "tests.test_pre.test_case",
+            "name": "test_case",
+            "status": "passed",
+            "start": 1000,
+            "stop": 2000,
+            "labels": [],
+            "parameters": [],
+            "attachments": [],
+        }))
+
+        respx.post("https://charisma.test/api/v1/launches").mock(
+            return_value=httpx.Response(201, json={"launchId": "launch-pre", "projectId": "proj-uuid"})
+        )
+        results_route = respx.post(
+            "https://charisma.test/api/v1/launches/launch-pre/results"
+        ).mock(return_value=httpx.Response(200, json={"accepted": 1}))
+        respx.post("https://charisma.test/api/v1/launches/launch-pre/close").mock(
+            return_value=httpx.Response(200, json={})
+        )
+
+        monkeypatch.setenv("CHARISMA_ENDPOINT", "https://charisma.test")
+        monkeypatch.setenv("CHARISMA_TOKEN", "pre-token")
+
+        runner = CliRunner()
+        result = runner.invoke(cli, [
+            "watch",
+            "--project", "pre-project",
+            "--results", str(results_dir),
+            "--", sys.executable, "-c", "import sys; sys.exit(0)",
+        ])
+
+        assert result.exit_code == 0
+        assert results_route.called
+        sent_results = json.loads(results_route.calls[0].request.content)["results"]
+        assert any(r["test_id"] == "hist_pre" for r in sent_results)
+        assert "results_sent=1" in result.output

{charisma_cli-0.1.6 → charisma_cli-0.2.0}/tests/test_watcher.py

@@ -733,8 +733,14 @@ class TestSettleScan:
             seen.add(key)
         assert len(seen) == 10
 
-    def test_settles_quickly_when_directory_empty(self, tmp_path: Path) -> None:
-        """An empty directory settles after one quiet period and enqueues nothing."""
+    def test_settles_quickly_when_directory_empty(self, tmp_path: Path, caplog) -> None:
+        """An empty directory settles after one quiet period and enqueues nothing.
+
+        Also asserts the settle path logs at INFO (not WARNING) — settling
+        cleanly is the normal case and must not raise a max_wait warning.
+        """
+        import logging
+
         results_dir = tmp_path / "allure-results"
         results_dir.mkdir()
 
@@ -742,19 +748,26 @@ class TestSettleScan:
         watcher = ResultsWatcher(str(results_dir), queue, _make_config())
 
         start = time.monotonic()
-        enqueued = watcher.settle_scan(quiet_period=0.3, max_wait=5.0, poll_interval=0.1)
+        with caplog.at_level(logging.INFO, logger="charisma_cli.watcher"):
+            enqueued = watcher.settle_scan(quiet_period=0.3, max_wait=5.0, poll_interval=0.1)
         elapsed = time.monotonic() - start
 
         assert enqueued == 0
         assert queue.empty()
         assert elapsed < 4.0  # returned on the quiet-period, not the max_wait cap
+        # Clean settle → INFO log, never a max_wait WARNING.
+        assert any(r.levelno == logging.INFO for r in caplog.records)
+        assert not any(r.levelno == logging.WARNING for r in caplog.records)
 
-    def test_respects_max_wait_when_directory_never_settles(self, tmp_path: Path) -> None:
+    def test_respects_max_wait_when_directory_never_settles(self, tmp_path: Path, caplog) -> None:
         """max_wait caps total time even if files keep arriving continuously.
 
         This documents the one bounded caveat: a directory that never goes quiet
-        stops at max_wait rather than blocking teardown forever.
+        stops at max_wait rather than blocking teardown forever. Also asserts a
+        WARNING is emitted so the up-to-max_wait teardown is observable rather
+        than silent.
         """
+        import logging
         import threading
 
         results_dir = tmp_path / "allure-results"
@@ -776,7 +789,8 @@ class TestSettleScan:
         writer.start()
         try:
             start = time.monotonic()
-            watcher.settle_scan(quiet_period=0.5, max_wait=1.5, poll_interval=0.1)
+            with caplog.at_level(logging.WARNING, logger="charisma_cli.watcher"):
+                watcher.settle_scan(quiet_period=0.5, max_wait=1.5, poll_interval=0.1)
             elapsed = time.monotonic() - start
         finally:
             stop.set()
@@ -784,6 +798,11 @@ class TestSettleScan:
 
         # Should stop at ~max_wait, not run indefinitely.
         assert 1.5 <= elapsed < 4.0
+        # Hitting max_wait must surface a WARNING (observability, not silent).
+        assert any(
+            r.levelno == logging.WARNING and "max_wait" in r.getMessage()
+            for r in caplog.records
+        )
 
     def test_scales_to_large_burst(self, tmp_path: Path) -> None:
         """A large number of files (simulating many workers/tests) is fully captured.