🐛 FIX: resolve regex vulnerabilities

[KyleKing]

Address confirm vulnerabilities from #143

Does not fully resolve the issue because there are unconfirmed/low risk vulnerabilities to consider and we don't know what was initially found that led to the ticket

[chrisjsewell]

Python already caches regex compilation, so it is not necessary to have any of these global variables

[KyleKing]

I know, but those are for the script in tests/test_redos.py to check them and compilation is on the µs scale compared to the overall overhead of Python + mdformat's architecture

We could remove that general check file entirely if you don't want to have it fuzz test moving forward, but the alternative to use lazy compilation while also fuzzing is to walk the AST, which makes that script much more complicated

TLDR: delete tests/test_redos.py and go back to lazy compilation?

[chrisjsewell]

Reviewed by checking the branch out and measuring the before/after behavior directly. The core fixes are correct and the suite is green (544 passed).

ReDoS fixes — confirmed real and resolved:

• texmath ```math block rules: the unpatched ^{3}math\s+?([^]+?)\s+?{3}grows super-linearly on `` ```math `` + a whitespace run with no closing fence — I measured **0.14 s / 1.1 s / 7.5 s** at 400 / 800 / 1600 chars. The rewritten pattern stays flat (**<3 ms** at 200k chars). Same class of fix applied toadmon`.
• The generic tests/test_redos.py guard discovers all 27 shipped patterns (including the ones nested in texmath's rules dict) and would catch a reverted texmath fix — verified.

Precompile cleanups (amsmath RE_OPEN, anchors, dollarmath label normalizer, tasklists): behavior-preserving, verified parity on valid/invalid inputs. Good hygiene to bundle in.

Three notes (details inline):

1. tests/test_admon.py — the new test_title_parsing_redos doesn't reproduce the blow-up; its input passes on the unpatched regex, so it doesn't guard the fix. Easy to strengthen.
2. texmath — the new capture trims surrounding whitespace / rejects whitespace-only content. Harmless, just flagging.
3. No CHANGELOG.md entry. Since this is the security-relevant fix (and #149 added one), it'd be good to record it here too, ideally noting which plugins/flavors were affected.

Nothing here blocks merging once the admon test is tightened. Nice work, and thanks for adding the categorical guard rather than just patching the two regexes.

---

Generated by Claude Code

[chrisjsewell]

Thanks @KyleKing — yeah, let's go back to lazy compilation.

To be clear, it's not the perf: you're right that the compile cost is µs-scale and immaterial next to Python startup + everything downstream. My concern is structural — I don't want the library's module layout (a set of new top-level re.Pattern globals) to be driven by what the test file can introspect. That's the test tail wagging the library dog. The actual security fix here is just the two pattern rewrites — texmath's ```math rules and admon's [^"]+ → [^"\s](?:[^"]*[^"\s])? — and those can stay inline where they lived. The rest of the hoisting (amsmath/anchors/dollarmath/tasklists) can come back out.

On tests/test_redos.py, though, I don't think it's a strict "keep the globals or delete it" — there are ways to keep fuzzing with lazy compilation:

• Behavioral (my preference): drive each plugin end-to-end — MarkdownIt().use(plugin) then md.render(<adversarial payload>) under a @pytest.mark.timeout. It doesn't care how or where the regexes are compiled, and it exercises the real attacker-controlled surface (the markdown source, plus the parser re-invoking rules across offsets — the O(n²) amplification you noted as Typo in Github "about" text #2). An isolated-regex check can both miss that and false-positive on patterns that adversarial input never actually reaches. The cost is maintaining a small payload per plugin.
• AST walk if you want the zero-maintenance automatic coverage — more code, but it fully decouples the guard from the module structure instead of forcing globals.

If either feels like too much for this PR, I'm also happy to drop test_redos.py here and land just the pattern fixes, then add a guard back separately. Your call on which you'd rather own — I lean behavioral since it maps to how the code is actually attacked.

---

Generated by Claude Code

[KyleKing]

I implemented the behavioral change, which I agree is a better approach overall and rolled back the now unnecessary eager compilation. I'll fix CI and then mark ready for re-review