stillwater-sc / universal / 25056252254

84%

master: 84%

Build Type

push

github

Committed by web-flow

Commit Message

fix(lns): per-algorithm tolerance for add/sub regression suite (#788)

* fix(lns): reorganizing the lns add/sub algorithm swap

* feat(lns): log-domain bound helper

* fix(lns): test suite cleanup post-reorganization (CodeRabbit)

Address CodeRabbit findings on PR #788 -- all in code from the
reorganization commit 7af6f4fb, none in the tolerance helper itself:

1. lns_test_suite.hpp: header comment said "blockfraction_test_suite.hpp"
   (copy-paste leftover). Fix to lns_test_suite.hpp.

2. lns_test_suite.hpp: out-of-range diagnostic messages all printed
   " * " regardless of the actual op. Fix per-verifier:
   - VerifyAddition uses " + "
   - VerifySubtraction uses " - "
   - VerifyMultiplication uses " * " (already correct)
   - VerifyDivision uses " / "

3. lns_test_suite.hpp: VerifySubtraction had an inconsistent early-return
   pattern (bailed at nrOfFailedTestCases > 0, returned literal 25)
   relative to the other three verifiers. Align to the standard
   nrOfFailedTestCases > 24 / return nrOfFailedTestCases pattern.
   Drive-by: VerifyDivision had the right bound but also returned
   literal 25; aligned for consistency.

4. log_add_algorithms.cpp: MANUAL_TESTING block referenced LNS4_2_sat
   which was not in scope (would fail to compile when MANUAL_TESTING=1).
   Add the local using-alias.

All 5 lns regression targets pass on gcc and clang.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* fix(lns): multiplication off-by-one + manual-test exit + canonical include (CodeRabbit)

Address CodeRabbit follow-up findings on PR #788:

1. lns_test_suite.hpp: VerifyMultiplication had an off-by-one early-return
   threshold (> 25 vs the > 24 used by Addition/Subtraction/Division), so it
   would collect one extra failure before bailing. Align to > 24.

2. log_add_algorithms.cpp: the MANUAL_TESTING block returned an unconditional
   EXIT_SUCCESS, which would hide failures from the test runner. Replace with
   the same conditional ... (continued)

Coverage Stats

115 of 155 new or added lines in 3 files covered. (74.19%)

22 existing lines in 4 files now uncovered.

45718 of 54230 relevant lines covered (84.3%)

6248193.84 hits per line