Fix the PMD violations reported by the Maven site (#122)

[anuragpaul602-netizen]

Closes #122.

The PMD report is now clean: 0 violations (PMD 7.7.0, ruleset pmd.xml), down from 89 in the issue (91 on my checkout — see the note on UnusedPrivateMethod below).

Breakdown

Count	Rule	Fix
68	UselessParentheses	Dropped the redundant outer parentheses in the MD4 rounds and helpers, the Envelopes SOAP builders, and the boolean/bitwise returns of WsmanClient, WinRMEndpoint, NTLMMessage, Type1Message, WmiHelper and ShellFileCopy.
11	UnnecessaryModifier	Removed public on interface methods and final on try-with-resources variables.
6	EmptyCatchBlock	Renamed the caught exception to ignored, the name PMD accepts for a deliberately ignored exception (allowExceptionNameRegex), keeping the existing explanatory comments. Type3Message's genuinely empty block also gained a comment: a missing SHA1PRNG leaves RND_GEN null and the constructor reports it.
2	AvoidBranchingStatementAsLastInLoop	NtlmCrypto.Cursor.skipUntil now loops on a new matchesAt(...) predicate instead of a labelled continue/return; WsmanClient.request accumulates into a Decoded sentinel and returns after the retry loop.
1	EmptyControlStatement	HttpTransport.readChunked reads the next trailer line in the loop body rather than in the condition.
1	UnusedFormalParameter	Removed the unused target parameter from the CipherGen constructor and its single call site.
2	UnusedPrivateMethod	Not in the issue's list — these two only surface when PMD's type resolution is degraded, but they are in the report on my checkout, and both are false positives on used methods. Renamed the private collectItems overload to collectRows so PMD resolves the call, and annotated RemoteDigest.matches, which PMD cannot see through Optional#get(), with a justified @SuppressWarnings.

The issue left open whether the light/* NTLM/crypto ports should get a targeted suppression instead. I went with the mechanical fix: those files are already reformatted to the project's style, so they are not byte-comparable with upstream anyway, and removing the parentheses is provably semantics-preserving. Happy to switch them to a class-level suppression if you would rather keep them close to the fork.

Tests

Everything above is behavior-preserving, and the NTLM paths (MD4, NtlmCrypto, WsmanClient.request, Envelopes) are covered end to end by the existing WsmanProtocolTest against FakeWsmanServer.

The one gap was the trailer-draining path rewritten in readChunked, which had no coverage. FakeWsmanServer gained a chunked-response mode (several chunks, a chunk extension, trailer fields after the terminating chunk) and WsmanProtocolTest a test that pages a WQL enumeration over it — a client that leaves the trailers in the socket desyncs the kept-alive NTLM connection and fails on the following request. I checked the test's sensitivity by deliberately truncating the drain, which makes it fail.

Verification

• PMD 7.7.0: 0 violations. Checkstyle: 0.
• SpotBugs: 36, down from 37 on main (the Type3Message empty catch), nothing new. Run with plugin 4.9.8.0 because the pinned version cannot read Java 25 bytecode on my machine.
• mvn test: 83 tests (was 82), all passing except WinRmCliTest.decodesCommandOutputUsingTheRemoteWindowsCodePage and WinRMWqlExecutorTest.testExecute, which fail identically on unmodified main here — Mockito's inline mock maker refuses to mock under JDK 25.
• Heads-up on the "mvn verify site still green" criterion: I could not confirm it locally, because SpotBugs and the site skin both break under JDK 25 (Unsupported class file major version 69, sun.misc.Unsafe.ensureClassInitialized) before and after these changes. It needs a CI run on a supported JDK.

CipherGen is a public class, so dropping a constructor parameter is technically a signature change; it is internal NTLM plumbing with a single caller, and I noted it in CHANGELOG.md under the in-progress 2.0.0 section. No README change: nothing here is visible to users of the documented API.

pmd:check is enabled in this PR (see the comment below), so the report cannot silently regress.

[anuragpaul602-netizen]

Pushed ce08ad0: pmd:check is now enabled, so the report cannot silently regress.

• Bound to the verify phase with the same pmd.xml ruleset the report uses, and printFailingErrors on so a failure names the rule and line.
• The plugin version is pinned to 3.26.0 — the version the parent manages for the site report — so the gate and the report always run the same PMD (7.7.0). Left unversioned, the build resolved 3.28.0 (PMD 7.17.0) and Maven warned about the missing version; the two would then drift apart on every plugin release.

Verified both directions: mvn verify passes on the branch, and re-introducing one UselessParentheses in MD4 fails it with PMD Failure: org.metricshub.winrm.light.MD4:203 Rule:UselessParentheses.

CPD is deliberately left alone — cpd-check would fail today, and that is a separate cleanup.

[bertysentry]

@codex please review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Bravo.

Reviewed commit: 04fd407ad3

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".