geotiff: parse_all_ifds raises on IFD chain cycle (#1913)

[brendancol]

Summary

Fixes #1913. parse_all_ifds in xrspatial/geotiff/_header.py silently truncated the IFD list on a cyclic chain. The loop condition was while offset != 0 and offset not in seen, so a repeat offset exited the loop and returned whatever had been parsed before the cycle. The other two malformed-chain branches in the same function (offset past EOF, chain over MAX_IFDS) raise ValueError, so the cycle case was the odd one out.

The fix restructures the loop to raise ValueError on a repeat offset with the same file is malformed wording. Callers now get a clear error rather than a half-parsed file that looks plausible.

Changes

• xrspatial/geotiff/_header.py: cycle branch now raises ValueError. Comment explains why (consistency with the sibling errors).
• xrspatial/geotiff/tests/test_header.py: existing TestIFDChainLoop asserted the old silent-truncation behaviour; updated to assert the new raise (two-IFD cycle and self-loop).
• xrspatial/geotiff/tests/test_ifd_cycle_1913.py: new test file covering cycle rejection on little-endian and big-endian, error message content, and regression guards for the two sibling malformed-chain branches and the normal multi-IFD chain.

Test plan

• New cycle tests pass: pytest xrspatial/geotiff/tests/test_ifd_cycle_1913.py -q (8 passed)
• Updated TestIFDChainLoop tests pass: pytest xrspatial/geotiff/tests/test_header.py -q
• Existing IFD chain cap tests still pass: pytest xrspatial/geotiff/tests/test_ifd_chain_cap.py -q
• Full geotiff suite green except for two pre-existing failures unrelated to this change (test_predictor2_big_endian_gpu_1517.py::test_gpu_predictor2_big_endian_int32_tiled_reproducer monkeypatch target missing, test_size_param_validation_gpu_vrt_1776.py::test_tile_size_positive_works tile_size mismatch). Both reproduce on main at b4f0eee.

[Copilot]

Pull request overview

This PR hardens the GeoTIFF IFD-chain parser by treating cyclic IFD chains as malformed input and raising a ValueError, aligning cycle handling with existing malformed-chain error branches (past-EOF and MAX_IFDS cap) and preventing callers from receiving a plausible-looking but truncated parse result.

Changes:

• Update parse_all_ifds to explicitly detect repeated IFD offsets and raise ValueError with “file is malformed” wording.
• Update existing cycle-related tests to assert the new raising behavior (including self-loop).
• Add a new targeted regression test module covering cycle detection (LE/BE), message content, and guarding the sibling malformed-chain branches plus normal chains.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
xrspatial/geotiff/_header.py	Makes IFD-chain cycles raise ValueError for consistent malformed-file handling.
xrspatial/geotiff/tests/test_header.py	Updates existing cycle-loop tests to expect the new exception-based contract.
xrspatial/geotiff/tests/test_ifd_cycle_1913.py	Adds regression coverage for cycle detection across endianness and for existing malformed-chain branches.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.