Add gcov/lcov coverage reporting and new test cases for payoffs, bonds, Heston, and cash flows

[tobydrinkall]

Summary

Adds CMake-based gcov/lcov coverage reporting infrastructure and 43 new Boost.Test cases targeting business-critical coverage gaps identified via baseline measurement.

Coverage infrastructure

New CMake option QL_ENABLE_COVERAGE=ON adds -fprofile-arcs -ftest-coverage flags (GCC/Clang only). Custom make coverage target runs: lcov --zerocounters → ctest → lcov --capture → lcov --extract .../ql/* → genhtml.

cmake -B build-cov -DQL_ENABLE_COVERAGE=ON -DCMAKE_BUILD_TYPE=Debug -DBOOST_ROOT=/usr -DQL_BUILD_EXAMPLES=OFF
cmake --build build-cov -j$(nproc)
make -C build-cov coverage
# → build-cov/coverage_report/index.html

Coverage results (before → after)

Metric	Before	After	Delta
Lines	72.2% (62,418/86,443)	72.6% (62,747/86,456)	+0.4% (+329 lines)
Functions	70.2% (11,083/15,785)	70.6% (11,140/15,788)	+0.4% (+57 functions)

Targeted file improvements:

File	Before	After
instruments/payoffs.cpp	27%	66.9%
pricingengines/bond/bondfunctions.cpp	47%	84.9%
cashflows/duration.cpp	0%	80.0%
cashflows/cashflows.cpp	64%	81.7%
processes/hestonprocess.cpp	60%	72.0%

New test files (4 files, 43 test cases)

test-suite/payoffs.cpp — PayoffTests (9 cases): NullPayoff, PlainVanillaPayoff (call/put/boundary), CashOrNothingPayoff, AssetOrNothingPayoff, GapPayoff, SuperFundPayoff, SuperSharePayoff, PercentageStrikePayoff, FloatingTypePayoff. Covers all payoff operator()/accept()/description() paths.

test-suite/bondfunctions.cpp — BondFunctionsTests (12 cases): date inspectors (startDate, maturityDate, isTradable), cash flow inspectors (nextCashFlowDate/Amount, previousCouponRate), accrual inspectors, duration (Simple/Macaulay/Modified + relationship modified ≈ macaulay/(1+y)), convexity, bps, basisPointValue/yieldValueBasisPoint, clean/dirty price from yield, yield-from-price round-trip, atmRate, zSpread, zero-coupon bond, Duration::Type operator<<.

test-suite/hestonprocess.cpp — HestonProcessTests (10 cases): accessors, size()/factors() for QE vs Broadie-Kaya, initialValues(), drift() with positive/negative variance across PartialTruncation/FullTruncation/Reflection, diffusion() matrix structure, apply() (log-normal spot + additive variance), evolve() with QE and PartialTruncation, all 5 standard discretization schemes.

test-suite/cashflowsmore.cpp — CashFlowsMoreTests (12 cases): CashFlows::npv with InterestRate (monotonicity in rate), bps, duration (all 3 types), convexity, basisPointValue/yieldValueBasisPoint, zSpread round-trip, atmRate, yield-from-NPV round-trip, SimpleCashFlow leg, previous/next cash flow iteration, NPV with YieldTermStructure and z-spread consistency.

Local verification

100% build, 1373 test cases passed, 0 tests failed (285.94s)

No CI on this fork — all verification is local."

Link to Devin session: https://app.devin.ai/sessions/531a528a8c414c328b721bf32799c28b
Requested by: @tobydrinkall

---

Devin Review

Status	Commit
⚪ Not started	—

Run Devin Review

---

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring

[devin-ai-integration]

Devin Review found 2 potential issues.

View 4 additional findings in Devin Review.

[devin-ai-integration]

🟡 BOOST_CHECK_CLOSE used with 0.0 expected value — wrong assertion for zero comparisons

Throughout test-suite/payoffs.cpp, BOOST_CHECK_CLOSE is used 17 times with an expected value of 0.0 (e.g., BOOST_CHECK_CLOSE(call(80.0), 0.0, 1e-10)). BOOST_CHECK_CLOSE performs a relative tolerance comparison, and Boost documentation explicitly states both arguments must be non-zero. While the tests happen to pass when both values are exactly 0.0 (due to safe_fpt_division internals), these assertions cannot properly catch bugs: if the payoff implementation were changed to return a tiny non-zero value (e.g. 1e-20) instead of exact 0.0, BOOST_CHECK_CLOSE would always fail regardless of the tolerance — because the relative difference against 0.0 is infinite. The existing test suite correctly uses BOOST_CHECK_SMALL or BOOST_CHECK_EQUAL for zero comparisons (e.g., test-suite/bondfunctions.cpp:314, test-suite/integrals.cpp:355). These 17 occurrences should use BOOST_CHECK_EQUAL(x, 0.0) (for exact zero) or BOOST_CHECK_SMALL(x, tol) (for approximate zero).

Prompt for agents

In test-suite/payoffs.cpp, all 17 uses of BOOST_CHECK_CLOSE with a 0.0 expected value should be replaced with BOOST_CHECK_EQUAL(x, 0.0) since the payoff results are exact zeros from std::max. The affected lines are: 48, 49, 55, 56, 76, 77, 82, 95, 96, 100, 112, 118, 133, 134, 148, 149, 164. For example, line 48 should change from BOOST_CHECK_CLOSE(call(100.0), 0.0, 1e-10) to BOOST_CHECK_EQUAL(call(100.0), 0.0). Similarly, in test-suite/hestonprocess.cpp line 177, BOOST_CHECK_CLOSE(diff[0][1], 0.0, 1e-10) should become BOOST_CHECK_EQUAL(diff[0][1], 0.0) since the diffusion matrix element is explicitly assigned 0.0.

---

Was this helpful? React with 👍 or 👎 to provide feedback.

[devin-ai-integration]

Fixed in 6b4bd0c — all 17 BOOST_CHECK_CLOSE(x, 0.0, tol) replaced with BOOST_CHECK_EQUAL(x, 0.0) since these payoff results are exact zeros from std::max. All 1373 tests still pass.

[devin-ai-integration]

🟡 BOOST_CHECK_CLOSE used with 0.0 expected value in diffusion matrix test

Same issue as in payoffs.cpp: BOOST_CHECK_CLOSE(diff[0][1], 0.0, 1e-10) uses relative comparison against zero. The diffusion matrix at ql/processes/hestonprocess.cpp:99 explicitly sets tmp[0][1] = 0.0, so the result is exact zero. BOOST_CHECK_EQUAL should be used instead, consistent with the existing test suite's patterns.

Suggested change

	BOOST_CHECK_CLOSE(diff[0][1], 0.0, 1e-10);
	BOOST_CHECK_EQUAL(diff[0][1], 0.0);

---

Was this helpful? React with 👍 or 👎 to provide feedback.

[devin-ai-integration]

Fixed in the same commit (6b4bd0c) — changed to BOOST_CHECK_EQUAL(diff[0][1], 0.0) since the diffusion matrix element is explicitly assigned 0.0.