forcedotcom / forcedotcom/code-analyzer-core

5.16.0 silently drops custom PMD rulesets that point at a sibling file by relative path

未关闭
#508 2 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看
主要语言
Java
星标
4
派生
6
平均合并
1 天 19 小时
30 天内合并 PR
9

描述

## What you see

After upgrading to `@salesforce/plugin-code-analyzer` 5.16.0, a scan that used to run our custom PMD rules stops running them, and nothing fails. The exit code is still 0. The only sign is one Critical violation attached to no file, with a summary like `Found 1 violation(s) across 0 file(s)`. A CI gate that does not use `--severity-threshold` reads green while zero Apex rules ran.

`sf code-analyzer config` and `sf code-analyzer rules` show the cause: `Cannot resolve rule/ruleset reference 'scanner/.xml/'`.

## Why it happens

Our repo has two custom rulesets: a full one that defines the rules, and a smaller one that reuses some of them with ``, a path relative to the repo root. Up to 5.15.x, PMD resolved that path against the repo being scanned, because the PMD process was started from there.

Core PR #503 (the CWE-427 hardening) now starts the PMD process from the engine's own install folder (`cwd: __dirname`). The PR calls this behaviour-preserving because every argument handed to `java` is an absolute path. That is true for the paths the engine passes, but not for paths written inside a custom ruleset: PMD still resolves those from the process's working directory, which is now the wrong folder. The whole ruleset then fails to load, and the PMD engine drops out of the run.

The 5.16.0 release notes do not mention it.

## How to reproduce

1. `scanner/full.xml` defines `MyRule`. `scanner/subset.xml` contains ``.
2. `code-analyzer.yml` in the repo root lists both files under `engines.pmd.custom_rulesets`.
3. Run `sf code-analyzer rules --rule-selector pmd:MyRule`. On 5.15.x the rule is listed. On 5.16.0 the ruleset fails to load with the error above.

Seen with plugin 5.16.0 (code-analyzer-core 0.53.0, pmd-engine 0.46.0, PMD 7.26.0) on macOS. The same configuration loaded cleanly on 5.14.0 and on the release before 5.16.0.

## What would fix it

Any one of these:

- Resolve paths inside custom rulesets against the config file's folder or the workspace root. One way: keep the working directory pinned, but add the workspace root to PMD's classpath, so `scanner/...` resolves as a classpath resource.
- Or state in the `custom_rulesets` docs and the 5.16.0 release notes that a custom ruleset must be self-contained (no relative `` to another file).

## Workaround

Generate the smaller ruleset as a self-contained copy of the rules it needs, instead of pointing at the full one. That is what we now do.

贡献指南

打开贡献指南

调研方向

Start with the PMD process setup described in Core PR #503 and reproduce using scanner/full.xml, scanner/subset.xml, and code-analyzer.yml. Run `sf code-analyzer rules --rule-selector pmd:MyRule` against the sample configuration on 5.16.0. Done means a relative rule reference loads from the repository configuration and `MyRule` is listed without the ruleset-resolution error.

由索引模型根据 Issue 内容生成。

评估

技术栈
java
领域
tooling
Issue 类型
缺陷
难度
4/5
预计耗时
3-5 天
活跃度
活跃
描述清晰度
基本清楚
新手友好度
55/100

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。