From 7f8a8829f9ac94b841f119eb7cce3487ef0e1df1 Mon Sep 17 00:00:00 2001
From: Dai Ha <dai.ha@magnolia-cms.com>
Date: Sat, 12 Sep 2026 12:58:26 +0700
Subject: [PATCH] fleetd #529 follow-up: the new helper's javadoc claimed a
 reach it does not have
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

CapturedLog's class javadoc said it is "the one way to pin or capture a logger's
level and output in this test tree". Measured on main at af95897, that is false:
nine test files still hand-roll the ListAppender + setLevel + finally
detachAppender pattern, with 42 setLevel calls on a raw logback Logger between
them.

None of those nine is a defect. Every one pairs its pin with a restore, so none
is the fleetd #525 leak, and #529's scope was the 19 unrestored pins only. The
problem is the sentence, not the code: a reader who believes "the one way" and
then greps finds nine counter-examples and cannot tell a leftover from a
violation. That is the same shape as a wrong reason in a comment — the text
survives while the fact under it moves.

Replaced with what is actually true: new code must use the helper, the pattern
still exists elsewhere, and here is the list plus the two commands that
re-measure it. The paragraph says to delete itself once the first command comes
back empty, rather than to keep a count up to date.

Javadoc only. mvn -f fleetd/pom.xml test-compile exit 0.
---
 .../dev/ltms/fleet/testing/CapturedLog.java   | 26 ++++++++++++++++---
 1 file changed, 23 insertions(+), 3 deletions(-)

diff --git a/fleetd/src/test/java/dev/ltms/fleet/testing/CapturedLog.java b/fleetd/src/test/java/dev/ltms/fleet/testing/CapturedLog.java
index 9980b73..e445f7f 100644
--- a/fleetd/src/test/java/dev/ltms/fleet/testing/CapturedLog.java
+++ b/fleetd/src/test/java/dev/ltms/fleet/testing/CapturedLog.java
@@ -21,9 +21,29 @@ import java.util.List;
  * fork. try-with-resources makes "restored the appender but not the level" impossible to write,
  * because there is only one thing to close.
  *
- * <p>This is the one way to pin or capture a logger's level and output in this test tree — do not
- * hand-roll the {@code ListAppender} + {@code setLevel} + {@code finally detachAppender} pattern;
- * use {@link #at} or {@link #of} instead.
+ * <p>New code must use this rather than hand-rolling the {@code ListAppender} + {@code setLevel} +
+ * {@code finally detachAppender} pattern: use {@link #at} or {@link #of}. It is <em>not</em> yet
+ * the only instance of the pattern in this test tree, and the earlier wording here said it was —
+ * which would leave a reader who greps unable to tell a leftover from a violation.
+ *
+ * <p>Measured on main at af95897 (2026-09-12): nine test files still hand-roll it, with 42
+ * {@code setLevel} calls on a raw logback {@code Logger} between them — {@code
+ * FleetdStartupReportTest}, {@code GitHostShapeReportTest}, {@code MemberCredentialsGapReportTest},
+ * {@code MemberTrustModelReportTest}, {@code FleetHealthMonitorTest}, {@code
+ * ClaudeCodeLauncherTest}, {@code HerdrPeerLauncherAllowListWiringTest}, {@code
+ * HerdrPeerLauncherCharterTest} and {@code OpenCodeLauncherTest}. Every one of them pairs its pin
+ * with a restore, so none is the fleetd #525 leak and none was in fleetd #529's scope, which was
+ * the 19 <em>unrestored</em> pins only. They are unmigrated, not broken.
+ *
+ * <p>Re-measure with the two commands below, from the repo root. A file that appears in the first
+ * list and not the second still hand-rolls the pattern. When the first list comes back empty, this
+ * paragraph is spent and the sentence above can go back to saying "the one way" — delete the
+ * paragraph then rather than updating the count.
+ *
+ * <pre>{@code
+ * grep -rlE '\.setLevel\(' fleetd/src/test/java --include='*.java' | grep -v CapturedLog.java
+ * grep -rl 'CapturedLog' fleetd/src/test/java --include='*.java'
+ * }</pre>
  */
 public final class CapturedLog implements AutoCloseable {
     private final Logger logger;