[hutiefang76]

Fixes MATH-1392: https://issues.apache.org/jira/browse/MATH-1392

Sample-size validation always reserves a column for an intercept, including when setNoIntercept(true) is used. A full-rank 2×2 design such as x = [[1, 2], [3, 4]], y = [5, 11] is therefore rejected before OLS or GLS can solve the uniquely determined coefficients [1, 2].

Count only the design columns actually used in both the flat and matrix entry points. Keep the existing default-intercept boundary, rejection of underdetermined and empty inputs, and covariance validation. This permits coefficient estimation for a square full-rank model; it does not make error variance estimable at zero residual degrees of freedom, and the Javadoc now states that limit.

Validation on the current head e406d597 (macOS arm64, JDK 17):

• All 60 OLS/GLS tests pass, with no failures, errors or skips. The original red run rejected the three square no-intercept cases; its other 57 cases passed.
• Root mvn -B clean verify passed across all 14 modules: 3,809 test cases, 33 skips, zero final failures or errors. The project's unchanged retry configuration recovered 51 randomized cases after 97 failed attempts; this was one root build invocation. Focused cases are included in this count.
• The three implementation/test files are byte-for-byte unchanged by the rebase onto master 73bf53b4. The changelog keeps both MATH-1678 and MATH-1392; the net PR still contains only its four intended files.
• git diff --check passed. Hosted CI for this rebased head is separate and remains to complete.

AI-generated description. DeepSeek and Codex assisted with investigation, implementation, tests and review. The Git state and validation figures were checked against the execution records.

[hutiefang76]

The three workflows for 6fc6ba7 are waiting for approval (action_required, zero jobs started). Could a maintainer approve them? Local mvn clean verify and the Java 8 OLS/GLS checks have passed; the PR description includes the existing automatic-retry results.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 87.19%. Comparing base (f554608) to head (e406d59).
⚠️ Report is 385 commits behind head on master.

Additional details and impacted files

@@             Coverage Diff              @@
##             master     #337      +/-   ##
============================================
+ Coverage     86.54%   87.19%   +0.64%     
+ Complexity     9787       91    -9696     
============================================
  Files           532      499      -33     
  Lines         35516    33514    -2002     
  Branches       6194     5848     -346     
============================================
- Hits          30738    29222    -1516     
+ Misses         3518     3169     -349     
+ Partials       1260     1123     -137     

☔ 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.

[aherbert]

Please rebase to resolve conflicts

[hutiefang76]

Rebased onto current master 73bf53b4; the new head is e406d597 with one linear commit. The only conflict was in changes.xml, where both MATH-1678 and MATH-1392 are retained. The three implementation/test files are byte-for-byte unchanged.

All 60 OLS/GLS tests pass. Root mvn clean verify also completes all 14 modules: 3,809 cases, 33 skips, no final failures or errors. The existing retry configuration recovered 51 randomized cases after 97 failed attempts; seeds, tolerances and retry settings are unchanged. Hosted CI still needs to complete on this head.

_{AI-generated reply; Git state and validation figures were checked against the execution records.}