lint: refuse an array element passed to a device routine that runs a seq loop

[sbryngelson]

Closes #1815

Enforces the workaround for #1815 so it cannot be reintroduced.

CCE OpenACC (19.0.0 through 21.0.2, -O2) miscompiles one pairing: a device routine that contains an acc loop at any level (seq included; the conforming loop vector in routine vector fails the same way), called with an array element as an actual argument. The element is read as garbage and never written back; either ingredient alone is fine, and CCE OpenMP offload of the same shape is correct. The pairing is what #1811 hit when it added Newton and RK4 loops to helpers that kernels called with q_prim_vf%vf(i)%sf(j,k,l) and blkmod1(k,l,q).

check_device_routine_element_args in lint_source.py: collect the GPU_ROUTINEs that contain a GPU_LOOP, close that set over device routines that call them (CCE inlines the chain, so the loop counts wherever it ends up), then flag any call to one of them, inside a GPU_PARALLEL_LOOP body or another device routine, whose argument is exactly an element reference such as a(i) or q(i)%sf(j,k,l). Expressions (a(i)/max(b(i), eps)), intrinsic and f_* calls, array sections (w(:,q)), whole arrays and host call sites pass. Contained procedures scope correctly.

Validation: master is clean; #1811 before its fix (c167a0aa) gets 38 findings in exactly the seven files that were changed by hand (m_rhs, m_hypoelastic, m_riemann_solver_hypo_hlld, m_acoustic_src, m_ibm, m_pressure_relaxation, m_variables_conversion); #1811 after its fix (1ee119b9) is clean. Four unit tests: flagged element in and out, flagged through a caller, host call site not flagged, scalars/expressions/loop-free routines not flagged.

One paragraph in contributing.md; the mechanism itself is documented in .claude/rules/common-pitfalls.md on #1811 and reproduced standalone in sbryngelson/compiler-bugs cce/acc-routine-element-by-reference.

[Copilot]

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Pull request overview

Adds a new source-lint rule (with tests + contributor guidance) to prevent reintroducing a known Cray OpenACC miscompile pattern: passing array elements into device routines that contain (or inline) acc loop seq-equivalent loops.

Changes:

• Implement check_device_routine_element_args in lint_source.py and wire it into the lint runner.
• Add unit tests covering flagged vs allowed call sites and propagation through device-routine call chains.
• Document the rule/workaround in docs/documentation/contributing.md.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 4 comments.

File	Description
toolchain/mfc/lint_source.py	Adds new linter logic to detect/flag element actual-args into “tainted” device routines and integrates it into main().
toolchain/mfc/test_lint_source.py	Adds unit tests exercising the new lint rule across host/device/kernels and call-chain propagation.
docs/documentation/contributing.md	Documents the new constraint and links it to the relevant issue + linter enforcement.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 62.26%. Comparing base (8010057) to head (3262cbb).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1816   +/-   ##
=======================================
  Coverage   62.26%   62.26%           
=======================================
  Files          84       84           
  Lines       21558    21558           
  Branches     3188     3188           
=======================================
  Hits        13423    13423           
  Misses       5937     5937           
  Partials     2198     2198           

☔ 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.