fix: bind TQQQ research input proof - #227
Conversation
Co-Authored-By: Codex <noreply@openai.com>
|
@codex review Exact head: |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b98a9ea2db
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".
| members: dict[str, bytes] = {} | ||
| total_size = len(manifest_raw) | ||
| for name in OUTPUT_FILENAMES: | ||
| raw = _read_member_from_root(inner_fd, name, _PROOF_TOTAL_BYTE_LIMIT - total_size) |
There was a problem hiding this comment.
Enforce the inner snapshot total-byte cap
For a crafted but hash-consistent proof, this budget allows the enclosed snapshot to grow to nearly 19 MiB whenever the outer manifest is small, because it subtracts len(manifest_raw) from the 19 MiB combined constant rather than independently limiting inner members to _MAX_TOTAL_MEMBER_BYTES (18 MiB). For example, a valid prices.csv near its 16 MiB limit plus three padded JSON files near 1 MiB each can pass this proof reader and _verify_tqqq_r1_snapshot_members, even though verify_tqqq_r1_snapshot rejects the same snapshot for exceeding its total-size contract. Track the inner total separately so proof verification preserves the existing snapshot resource boundary.
Useful? React with 👍 / 👎.
|
Closing unmerged under the frozen F1 phase contract: exact-head Codex review delivered a current P2/unresolved thread; PR-stage remediation is prohibited. |
Implements the three frozen F1 proof bindings and the required QPK compatibility pin update.\n\nLocal gates: focused 102 passed; full 621 passed, 3 skipped; ruff and compileall passed.\n\nBase: 9dd3077. Head: b98a9ea.