[ViralBShah]

Follows a review of the package for structure and soundness. This PR set the version to 0.7.0, but these changes are to be released as 0.6.1, in a follow-up that also stops exporting solve!.

Changes to the accepted element types are in #60, so klu, klu! and ldiv! accept the same types after this PR as before it.

Behaviour changes

• Julia 1.12 is now required. SparseArrays and LinearAlgebra compat move with it. Test stays at 1.10, because that stdlib is still v1.11.0 on Julia 1.12 and 1.13.
• Failed factorizations release their factors. After klu! or klu_factor! throws, K \ b, rcond, condest and det throw until a factorization succeeds.
• Fields are const. Only _symbolic, _numeric, _perms and nzval can still be assigned.
• An earlier failure no longer blocks a retry. klu_factor! used to rethrow a stale error, even after a temporary lack of memory.
• solve! with check=false no longer throws when it factors on demand.
• klu rejects empty matrices with a clear ArgumentError, as lu does.
• K.symbolic and K.numeric are no longer listed by propertynames(K), because they hold raw pointers into libklu's memory. They can still be accessed.

Memory safety

• Out-of-bounds reads: a factorization shares vectors with the caller, so dropzeros!(A) or resize! could leave K.nzval too short, and libklu read past its end. Operations that read these vectors now check the lengths first. Sharing itself is unchanged, to avoid using more memory, and the docs now say what the caller must not do.
• Shared pattern vectors: the pattern check in klu! shifted colptr and rowval in place, which other factorizations sharing them could see while inside libklu.
• Permutations: klu_analyze! validated the caller's P and Q, then used them, so another task could change them in between. It now validates and uses copies.
• Pointer assignment: assigning the current _numeric or _symbolic pointer freed it and left it dangling. Replacing _symbolic now also releases the numeric factors that depend on it.

Memory leak

deserialize doesn't call the constructor, so a deserialized factorization had no finalizer and never freed the libklu objects it made on demand. The finalizer is now registered when the first of them is attached.

Peak memory growth over 60 rounds of solving with a 300,000 × 300,000 tridiagonal matrix and running GC:

	Before	After
Ordinary factorization	246 MiB	246 MiB
Deserialized factorization	2892 MiB	266 MiB

Failure paths and locking

• Wrong answers after a failure: when klu! stopped partway it left a mix of old and new factors, and a failed klu_factor! kept the earlier factors, so K \ b returned finite wrong answers. rcond and condest could then also make issuccess return true.
• Unlocked reads: issuccess, nnz and show now read under the lock, and show prints factors from one factorization.
• klu_sort: its result is now checked.
• klu_refactor! dropped keyword arguments.

Other changes

• init_suitesparse is now called from _common rather than __init__, as the stdlib solvers do on entry.
• Adds logabsdet, and so det and logdet, including for K' and transpose(K).
• SingularException reports a column with no pivot, and KLU_INVALID has a clearer message.
• ldiv! accepts a single row with any column stride.
• The docs for the scale setting now describe rows rather than columns.
• Removes the AdjointFact and TransposeFact aliases and the unused @isok macro.
• Adds Serialization as a test dependency.

Testing

On Apple silicon macOS, the test suite passes on Julia 1.12.7 and 1.13.3, and Aqua and the docs build pass on 1.12.7. Linux and Windows are left to CI.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.29%. Comparing base (146612b) to head (260e88d).

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #59      +/-   ##
==========================================
- Coverage   97.31%   97.29%   -0.03%     
==========================================
  Files           1        1              
  Lines         373      407      +34     
==========================================
+ Hits          363      396      +33     
- Misses         10       11       +1     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

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