[Duo-ee]

JenkinsLookup3.decode writes the payload with out.view("uint8")[:] = b[:-4], so an output buffer larger than the decoded payload silently broadcasts the single byte across the whole buffer:

codec = JenkinsLookup3()
out = np.zeros(4, dtype="u1")
codec.decode(codec.encode(b"x"), out=out)
out.tobytes()  # b'xxxx'

A buffer with the same byte count but another dtype/shape raises a broadcasting error, and a bytearray raises AttributeError.

This compares memoryview(out).nbytes with the payload before writing anything, and fills the buffer through a flat uint8 view:

j = JenkinsLookup3()
enc = j.encode(b"x")
out = np.zeros(4, dtype="u1")
j.decode(enc, out=out)  # ValueError: Output buffer has 4 bytes, but the decoded payload has 1 bytes.

The buffer is left untouched when the sizes differ, and a same-size bytearray or a multi-dimensional array now decodes as expected.

Fixes #894.

Tests: tests/test_jenkins.py gains test_out_too_large_is_rejected, test_out_with_other_dtype_is_filled and test_out_bytearray_is_filled.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.74%. Comparing base (4afa628) to head (f6473ba).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #895      +/-   ##
==========================================
+ Coverage   89.71%   89.74%   +0.03%     
==========================================
  Files          27       27              
  Lines         914      917       +3     
==========================================
+ Hits          820      823       +3     
  Misses         94       94              

Files with missing lines	Coverage Δ
src/numcodecs/checksum32.py	100.00% <100.00%> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Duo-ee]

This duplicates #891, which already fixes #894 (and #890), including bytearray, C/F-ordered matrices and strided destinations. I missed it because #894 does not list #891 under closedByPullRequestsReferences — opening this PR was redundant.

Please treat #891 as the fix; this branch can be closed.

[Duo-ee]

Same note as on the sibling PRs: the red build jobs fail during collection with ValueError: invalid literal for int() with base 10: 'post0' from the zfpy 1.0.1.post0 version string, on every 3.12–3.14 job of every open PR in this repo. The change itself (and #896/#897) compiles and passes wherever zfpy is absent.