[blink1073]

PYTHON-6188

Changes in this PR

The nightly fuzz run fails because a document containing a Regex can compare unequal between the C extension and pure Python even though both decoded it correctly. Corpus fuzzing also ran only on the nightly, so a change that broke a seed would not fail PR CI.

• Fix the comparison so equivalent Regex values compare equal, and commit the failing document as a regression reproducer.
• Replay the full seed corpus in the PR fuzz-regression task, not just the committed reproducers; fail fast when the corpus is missing or empty.

Test Plan

• The committed reproducer fails before the fix and passes after.
• After the fix, all three fuzz targets replay the full 485-seed corpus with zero divergences; the corpus previously stopped the nightly at seed 38.

Checklist

Checklist for Author

• Did you update the changelog (if necessary)? (not needed: the change is dev-only fuzz and CI tooling)
• Is there test coverage?
• Is any followup work tracked in a JIRA ticket? (none)

Checklist for Reviewer

• Does the title of the PR reference a JIRA Ticket?
• Do you fully understand the implementation? (Would you be comfortable explaining how this code works to someone else?)
• Is all relevant documentation (README or docstring) updated?

[Copilot]

🟢 Approval recommended

The reviewed changes address the reported regex divergence and ensure full corpus coverage in PR fuzz regression runs.

No open issues 🎉

What changed in this PR

This pull request fixes false cross-implementation divergences for BSON Regex values and expands PR fuzz regression runs to replay the full generated seed corpus deterministically.

Changes:

• Canonicalize Regex patterns and flags for consistent comparisons.
• Generate and expose the seed corpus during regression runs.
• Replay committed reproducers and all corpus seeds, failing fast for missing or empty inputs.
• Document the expanded PR fuzz behavior.

File	Description
fuzz/​README.md	Documents deterministic full seed-corpus replay during PR regression tasks.
fuzz/​fuzz_shared.py	Normalizes equivalent C-extension and pure-Python Regex values consistently.
.evergreen/​scripts/​setup_tests.py	Builds and exposes the seed corpus for all fuzz tasks, including regression runs.
.evergreen/​scripts/​run_tests.py	Validates and replays regression inputs plus the full seed corpus for each fuzz target.

_{Review effort: Lite}

[NoahStapp]

Some comment and documentation suggestions for clarity, otherwise looks good!

[NoahStapp]

What does it mean for an input to "mutate" in this context?

[blink1073]

As in static, without the random fuzzing (mutating).

[NoahStapp]

Could you rephrase to "random fuzzing" instead?

[blink1073]

Yep, it wouldn't let me do it in the UI, in progress