Name the failing test, its error and its job link instead of a dash

Author: aidangarske

.github/scripts/ci_triage.py

@@ -272,6 +272,27 @@ def job_unit(name):
     return target, (m.group(1) or m.group(2)) if m else "unknown"
 
 
+HOST_JOB_RE = re.compile(r"^host / (?:Run|Build) / (.+) \((.+)\)$")
+
+
+def job_urls(run_id):
+    """(unit, ref) -> the URL of the job that ran it.
+
+    Linking the run only makes a reader hunt through 250 jobs for the red one.
+    """
+    out = {}
+    for j in all_jobs(run_id):
+        name, url = j.get("name", ""), j.get("html_url")
+        m = HOST_JOB_RE.match(name)
+        if m:
+            out[(m.group(1), m.group(2))] = url
+            continue
+        t, ref = job_unit(name)
+        if t:
+            out[(t, ref)] = url
+    return out
+
+
 def classify_jobs(run_id, attempt, retry_note):
     def by_unit(jobs):
         out = defaultdict(list)
@@ -814,6 +835,7 @@ def report(attempt, retry_note):
     issues = {store.parse_fp(i.get("body", "")): i for i in open_issues()}
     issues.pop(None, None)
 
+    urls = job_urls(RUN_ID)
     total = len(all_jobs(RUN_ID))
     prev = previous_job_count()
     close_allowed = not (prev and total < 0.5 * prev)
@@ -870,10 +892,26 @@ def report(attempt, retry_note):
             "failing": fail_refs.get(eid, []) if observation != store.PASS else [],
             "log": logs.get(eid, "")[-LOG_TAIL_CHARS:],
         }
+        # The real failing target + its error + a link to the job that ran it.
+        # This used to be a hardcoded "-" profile column and a bare ":x: fail",
+        # which told a reader nothing they could act on.
         rows = [
-            {"profile": "-", "ref": ref, "status": ":x: fail"}
-            for ref in fail_refs.get(eid, [])
-        ] or [{"profile": "-", "ref": "all", "status": ":white_check_mark: pass"}]
+            {
+                "test": r.get("target") or r.get("stage", "build"),
+                "ref": r.get("ref", "?"),
+                "detail": (r.get("detail") or "").strip(),
+                "url": urls.get((eid, r.get("ref"))),
+            }
+            for r in results
+            if r["id"] == eid and r["status"] == "fail"
+        ]
+        # cross targets emit no results rows, so fall back to the job itself
+        if not rows and observation != store.PASS:
+            rows = [
+                {"test": eid, "ref": ref, "detail": "job failed",
+                 "url": urls.get((eid, ref))}
+                for ref in fail_refs.get(eid, ["unknown"])
+            ]
         # collected, not read back from `issues`: a dir filed for the first time
         # tonight is not in that map, and its card entry would have no link
         filed[eid] = apply(eid, observation, issue, verdicts.get(eid), state, rows,

.github/scripts/issue_store.py

@@ -111,34 +111,46 @@ def render_body(eid, state, ai, failing_rows, run_url, now):
         f"<!-- ci-fp: v1|{eid} -->",
         f"<!-- ci-state: {json.dumps(marker, sort_keys=True)} -->",
         "",
-        f"**`{eid}`** is failing in the nightly. "
-        f"{state.get('streak', 1)} consecutive night(s), since {state.get('first', now)}.",
+        f"**`{eid}`** is failing in the nightly — "
+        f"{state.get('streak', 1)} night(s) running, since {state.get('first', now)}.",
         "",
-        "| profile | wolfSSL ref | status |",
+    ]
+
+    if ai:
+        lines += [
+            f"### {ai.get('severity', '?')} · {ai.get('category', '?')} · "
+            f"affects {ai.get('blast', '?')}",
+            "",
+            f"- **Symptom** — {scrub(ai.get('symptom', ''))}",
+            f"- **Likely cause** — {scrub(ai.get('cause', ''))}",
+            f"- **Next** — {scrub(ai.get('next', ''))}",
+            "",
+        ]
+
+    # "wolfSSL" not "wolfSSL ref": the column answers "which wolfSSL broke it",
+    # and master-vs-stable is the single most useful fact here -- a stable-only
+    # break is a released bug, a master-only one is upstream drift.
+    lines += [
+        "### What failed",
+        "",
+        "| test | wolfSSL | error |",
         "|---|---|---|",
     ]
     for row in failing_rows:
-        lines.append(f"| {row['profile']} | {row['ref']} | {row['status']} |")
+        test = f"`{row['test']}`"
+        if row.get("url"):
+            test = f"[{test}]({row['url']})"
+        detail = (row.get("detail") or "").replace("|", "\\|")[:120] or "—"
+        lines.append(f"| {test} | `{row['ref']}` | {detail} |")
     lines.append("")
 
-    if ai:
+    if ai and ai.get("upstream"):
         lines += [
-            f"**Severity:** {ai.get('severity', '?')} · "
-            f"**Category:** {ai.get('category', '?')} · "
-            f"**Blast radius:** {ai.get('blast', '?')}",
-            "",
-            f"**Symptom** — {scrub(ai.get('symptom', ''))}",
-            f"**Likely cause** — {scrub(ai.get('cause', ''))}",
-            f"**Next** — {scrub(ai.get('next', ''))}",
+            "> Fails only against wolfSSL `master` and passes on stable — this "
+            "is more likely a wolfSSL regression than an example bug. Consider "
+            "filing against wolfSSL/wolfssl.",
             "",
         ]
-        if ai.get("upstream"):
-            lines += [
-                "> Fails only against wolfSSL `master` and passes on stable — this "
-                "is more likely a wolfSSL regression than an example bug. Consider "
-                "filing against wolfSSL/wolfssl.",
-                "",
-            ]
 
     if state.get("log"):
         lines += [