[None][fix] Keep sysinfo distro probe working without the distro module; fail empty test-list renders loudly

jenkins/L0_Test.groovy

@@ -4035,6 +4035,24 @@ def renderTestDB(pipeline, testContext, llmSrc, stageName, preDefinedMakoOpts=nu
         }
     }
 
+    // Log the resolved mako match on every render, tagged with stage+context.
+    // transformMakoArgsToJson already echoes the bare JSON ("Test DB Mako opts:"),
+    // but unlabeled and far upstream of the render; the preDefinedMakoOpts path
+    // skips it entirely. This co-locates the match with the stage/context and the
+    // "-> N tests" summary below under one greppable renderTestDB: prefix, so a
+    // wrong-but-non-empty render (a stale/unexpected sysinfo value selecting the
+    // wrong block) is diagnosable per stage, not just the "na"/empty cases.
+    echo "renderTestDB: stage=${stageName} context=${testContext} mako match: ${makoOpts}"
+
+    if (makoOpts.contains('"na"')) {
+        // "na" is a sysinfo probe failure sentinel (see get_sysinfo.py). Blocks
+        // conditioned on the failed property silently drop out of the render, so
+        // even a non-empty list may be missing tests. Warn here, where every
+        // sysinfo-based stage passes through, not just when the list ends up empty.
+        echo "WARNING: renderTestDB: some sysinfo probes returned \"na\": ${makoOpts}. " +
+             "Test-db blocks conditioned on those properties (e.g. linux_distribution_name: ubuntu*) " +
+             "will NOT be selected."
+    }
     sh "pip3 install --extra-index-url https://urm.nvidia.com/artifactory/api/pypi/sw-tensorrt-pypi/simple --ignore-installed trt-test-db==1.8.5+bc6df7"
     // CBTS Layer 3: download the pre-built cbts_test_db/ tarball that the
     // orchestrator uploaded to Artifactory (see getCbtsResult in
@@ -4081,10 +4099,30 @@ def renderTestDB(pipeline, testContext, llmSrc, stageName, preDefinedMakoOpts=nu
     ].join(" ")
 
     sh(label: "Render test list from test-db", script: testDBQueryCmd)
-    def testCount = sh(returnStdout: true, script: "wc -l < ${testList} | tr -d ' '").trim()
+    // Count non-empty lines, not newlines: trt-test-db writes the test names
+    // with no trailing newline, so `wc -l` undercounts by one -- it reports 0
+    // for a single-test render, which would trip the empty-list guard below
+    // (and mis-report every stage's count by one). `grep -c .` is agnostic to
+    // the missing terminator. It exits 1 for no matches (a legitimately empty
+    // render -> count "0"); accept only that, so a read failure (exit 2:
+    // missing file, unreadable, etc.) still aborts the step instead of being
+    // silently masked.
+    def testCount = sh(returnStdout: true, script: "grep -c . -- ${testList} || test \$? -eq 1").trim()
     def testDBLabel = (cbts != null && cbts.test_db_dir_override) ? "CBTS-narrowed [${cbts.scope}]" : "source"
     echo "renderTestDB: stage=${stageName} context=${testContext} test-db=${testDBLabel} dir=${testDBPath} -> ${testCount} tests"
     sh(script: "cat ${testList}")
+    if (testCount == "0") {
+        // An empty render is never legitimate here: every launched stage must
+        // have tests (CBTS drops stages with an empty selection before launch).
+        // Fail now with the match query rather than letting pytest --collect-only
+        // exit 5 later with an unattributable "Test collection failed" message.
+        def hint = makoOpts.contains('"na"') ?
+            " Some sysinfo probes returned \"na\" (see the match JSON above); a broken probe" +
+            " (e.g. the python 'distro' module missing) makes conditions like" +
+            " linux_distribution_name: ubuntu* match nothing." : ""
+        error("renderTestDB: rendered EMPTY test list for stage=${stageName} " +
+              "context=${testContext} test-db=${testDBLabel}. Match query: ${makoOpts}.${hint}")
+    }
     recordRenderedStageAttemptEstimate(pipeline, llmSrc, testList, stageName, testCount, clusterName)
 
     return testList

requirements-dev.txt

@@ -3,6 +3,9 @@
 # before pytest collection so already-imported Diffusers modules match the
 # package files. 0.39.0 matches the runtime floor in requirements.txt.
 diffusers==0.39.0
+# get_sysinfo.py's Linux-distribution probe. Must be explicit: it used to arrive
+# transitively via openai, and openai 3.3.1 dropped it, which silently emptied
+# every sysinfo-rendered CI test list (linux_distribution_name became 'na').
 distro==1.9.0
 boto3
 einops

tests/integration/defs/sysinfo/get_sysinfo.py

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2024-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-FileCopyrightText: Copyright (c) 2024-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
@@ -112,10 +112,28 @@ def get_linux_distribution():
     try:
         import distro
         return (distro.id(), distro.version(), distro.codename())
-    except:
-        logger.warning(
-            "Unable to use distro module, defaulting operating system to ('na', 'na', 'na')"
-        )
+    except ImportError:
+        # distro is not a direct test dependency historically; it used to arrive
+        # transitively (e.g. via openai<3.3.1) and silently disappeared when that
+        # dependency was dropped. Fall back to the stdlib rather than to 'na'.
+        distro_reason = "distro module not installed"
+    except Exception as e:
+        # Best-effort probe: distro imported but id()/version()/codename() raised.
+        # Any failure here should degrade to os-release, not crash the render, so
+        # the broad catch is deliberate; the reason is preserved for the log below.
+        distro_reason = f"distro probe raised {type(e).__name__}: {e}"
+        logger.warning(f"{distro_reason}; falling back to os-release")
+    try:
+        # Python 3.10+; reads /etc/os-release, same source distro uses.
+        os_release = platform.freedesktop_os_release()
+        return (os_release.get("ID", "na"), os_release.get("VERSION_ID", "na"),
+                os_release.get("VERSION_CODENAME", "na"))
+    except OSError:
+        logger.error(
+            f"Cannot determine the Linux distribution ({distro_reason}; "
+            "/etc/os-release also unreadable); reporting ('na', 'na', 'na'). "
+            "Test-db conditions matching linux_distribution_name (e.g. ubuntu*) "
+            "will select ZERO tests and the rendered test list will be empty.")
         return ("na", "na", "na")