[Beforerr]

In _read_vdr3/_read_vdr2, the pad-value offset was computed after dropping NOVARY dims, i.e. from the reduced dim count. On disk the VDR stores zDimSizes and DimVarys for every dim, so any zVariable with a NOVARY dim and a pad value got Pad read from inside the DimVarys array (e.g. -4294967296 instead of -7 for an int64 pad).

Fix: compute the offset from the stored zNumDims before dropping NOVARY dims.

Test: writes a 2-D zVariable with a pad value, flips one DimVary to NOVARY in the file bytes (cdfwrite cannot emit NOVARY zVariable dims), and checks varinq().Pad. Fails on main, passes here.

[Copilot]

🟡 Changes recommended

The CDF 2 parsing branch lacks targeted regression coverage.

1 open finding

• Add CDF 2 zVDR regression coverage for NOVARY pad parsing · New

What changed in this PR

Fixes pad-value parsing for zVariables containing NOVARY dimensions.

Changes:

• Calculates pad offsets before filtering NOVARY dimensions.
• Adds a CDF 3 regression test.

File	Description
cdflib/​cdfread.py	Corrects CDF 2/3 zVDR pad offsets.
tests/​test_cdfwrite.py	Tests CDF 3 NOVARY pad parsing.

🧠 Review effort: Balanced

---

💡 Configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[bryan-harter]

Thanks for finding this bug! Merging now, and it will go out in the next deployment.