feat(script): Verify checker labelling invariants (e.g., guideline:X ⇔ X:rule, profile:default ⊆ profile:sensitive ⊆ profile:extreme) with the label-tool
#4275
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
whisperity
added
enhancement 🌟
tools 🛠️
whisperity
added
the
label-tool 🔖
whisperity
changed the title
guideline:X ⇔ X:rule, default ⊆ sensitive ⊆ extreme) with the label-toolguideline:X ⇔ X:rule, profile:default ⊆ profile:sensitive ⊆ profile:extreme) with the label-tool
Automatically generate the `doc_url` labels for checkers for analysers which we know how to do this. (Currently implemented only for Clang SA and Clang-Tidy.) This tooling does a single HTTP request to download a "Table of Contents" (ToC) document and uses HTML DOM scraping to extract the list of checkers and their corresponding documentation link. There was existing prior work for this feature, but those scripts were about 3 years old (introduced in 2021. Nov, commit aa72dc0), and they ceased to properly work. For example, for _Clang-Tidy_, the contents of the ToC changed in a way that the previously used XPath expression did not match anything at all. In addition, the previous work at `doc_url.py` (later renamed `doc_url_generate.py`) only generated labels for the checkers that were _already_ present in the configuration file, defeating the purpose of using a script to determine new labels for checkers. This version reintroduces an improved DOM scraping logic, and enables generating `doc_url` for checkers that are not yet present in the configuration file, all the while using the new `label_tool/` package structure introduced in a previous commit.
Automatically generate the `severity` labels for checkers of analysers which we know how to. (Currently implemented only for the Clang diagnostic (warning) flags exposed through Clang-Tidy.) There was existing prior work for this feature, but that script was about 3 years old (introduced in 2021. Nov, commit 8d1a7fe), and due to changes in the DOM of Sphinx documents, ceased to work properly since. This refactoring is smaller in scope than the `doc_url` generation that was done in a previous commit, as the original `compiler_warnings.py` script accurately generated severities (and `doc_url`s, for that matter) for previously "unknown" "checkers" as well.
…`severity_generate`
Considering that all previously developed label-generating tools are now subsumed into the new `label_tool` package, the extra directory is unnecessary.
|
Added a rule to tear out |
whisperity
force-pushed
the
feat/script/label-tool/invariant-verification
branch
from
7210e00 to
414cee2
Compare
whisperity
added
the
RDY-OnHold 🛑
After upgrade on `master` to `pylint==3.2.4`, new warnings are enabled which were not honoured in `label-tool`.
In some cases it might be deemed necessary that a label (e.g., a higher or lower `severity`) is more appropriate than what the tool would otherwise auto-generate. In such cases, a `label-tool-skip:LABEL_KEY` (e.g., `label-tool-skip:severity`, as applied in commit 04d27ab) can indicate to stamp the fact that a developer overruled the tools' decisions. Prior to this commit, the tooling actually disregarded this information, but now I added the necessary handling to the existing aspects.
There were multiple occasions in the past at which point it was realised that the checker-labelling in the configuration files did not conform to some "invariants" which we all, or mostly, agreed upon to upkeep, but only did so ad-hoc and implicitly, without any tooling-based verification. N.B., violation of these invariants did not result in **serious** issues or bugs, but often caused weird and quirky behaviour which was non-trivial to understand for users. This patch extends the `label-tool` with facilities to verify and, if meaningful for the specific rule, automatically fix the label set of individual checkers as one entity (as opposed to the `doc_url_generate` or `severity_generate` tools, which only considered the value of a unique single label), upholding invariants **between** labels, e.g., `profile:` labels. This patch contains 4 implemented rules: - `profile:default` ⊆ `profile:sensitive` ⊆ `profile:extreme` - `guideline:sei-cert` ⇔ `sei-cert:<rule-number>` - `guideline:sei-cert` ⇒ `profile:security` - clangsa (alpha.* ∪ debug.*) ⇒ ¬( `profile:(default|sensitive|extreme|security|portability)` )
whisperity
force-pushed
the
feat/script/label-tool/invariant-verification
branch
from
414cee2 to
3a46b58
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Important
⛔ Blocked by #4274.
There were multiple occasions in which the labelling of the checkers was lacking some serious invariants which we all implicitly agreed upon, but never actually and thoroughly verified with tooling. Violating these invariants should have never resulted in serious issues, but quirky or unexpected behaviour could have occurred.
This patch extends the
label-toolwith facilities to verify, and, optionally, support automatically fixing the label set of each checker "globally" to uphold the following invariants:profile:default⊆profile:sensitive⊆profile:extremeguideline:sei-cert⇔sei-cert:<some rule's ID>guideline:sei-cert⇒profile:securityclangsa(alpha.*∪debug.*)⇒ ¬(profile:(default|sensitive|extreme|security|portability))