39. EN 50716 Annex A

39.1. Table A.1 — Lifecycle issues and documentation (5.3)

39.1.1. 1-5

Covered by the table of documents in section 5.3.2.4 of the EN 50716 norm mapping.

39.1.2. 6

Covered by

39.1.3. 7

Covered by Testing Plan.

39.1.4. 8

Covered by

39.1.5. 9

N/A; no architecture and no separate design specification is written.

The certified libraries have no overall architecture and a rather simple high-level design. Each module provides functions and data structures around a single well-defined topic. All modules have a doc-comment describing the design and contents of the module.

39.1.6. 10

N/A; no architecture and no separate design specification is written.

The certified libraries have no overall architecture and a rather simple high-level design. Each module provides functions and data structures around a single well-defined topic. All modules have a doc-comment describing the design and contents of the module.

39.1.7. 11

Covered by the Certified standard library API docs (function signatures and doc-comments).

39.1.8. 12

Covered by Integration tests.

39.1.9. 13

N/A; there is no hardware in the certification scope.

Ferrous Systems certifies the certified libraries as libraries, to be used in other systems whose requirements are unknown. Users of the certified libraries should consider their specific system safety requirements when developing safety related software with the certified libraries.

39.1.10. 14

Covered by Development Process and Safety Case.

39.1.11. 15-17

Covered by the module doc-comments (see Architecture and software design) and by Development Process.

39.1.12. 18

The source code and its configuration history are provided with the ferrocene-src component, see Configuration.

39.1.13. 19-20

Covered by

39.1.14. 21-24

Covered by Test results overview.

39.1.15. 25

Covered by Validation Process and Safety Case.

39.1.16. 26

Covered by Tool safety assessments and Tool Analysis.

39.1.17. 27

Covered by Ferrocene Release Notes.

39.1.18. 28-38

N/A; the certified libraries are not configurable by application data.

The certified libraries are delivered as pre-compiled libraries and have to be statically linked into applications (see LIBS_CSTR_0080_LINK_STATIC). There is no way to configure a certified library binary after it is compiled and linked statically.

39.1.19. 39-41

Covered on the Ferrocene side by Installation Procedures; the deployment into the target system is performed by the customer, see 9.1 of EN 50716.

39.1.20. 42-45

Covered by Patching process, Change Tracking and Known Problems Tracking.

39.1.21. 46-47

Written by the Assessor, see Safety Assessment.

39.2. Table A.2 — Software Requirements Specification (7.2)

39.2.1. 1

Not used; the requirements are written in natural language (doc-comments) together with the function signature, which is a formal and machine-checked description of the interface.

39.2.2. 2

Applied by the structure of doc-comments and the overall structure of the core library.

39.2.3. 3

Not used; the requirements are written in natural language (doc-comments) together with the function signature, which is a formal and machine-checked description of the interface.

39.3. Table A.3 — Software architecture (7.3)

39.3.1. 1

Covered by Programming language (memory safety, bounds checking, strong typing).

39.3.2. 2-4

Not used; the certified libraries are single-channel software without recovery, reconfiguration or diversity. Detecting and handling failures of the enclosing system is the responsibility of the user of the certified libraries.

39.3.3. 5

Applied by the language itself: out-of-range indexing, overflow and similar violations terminate in a documented panic, see Doc-comments in the certified libraries.

39.3.4. 6-14

Not used; the certified libraries are single-channel software without recovery, reconfiguration or diversity. Detecting and handling failures of the enclosing system is the responsibility of the user of the certified libraries.

39.3.5. 15

Covered by Failure analysis.

39.3.6. 16

Not used; the certified libraries are single-channel software without recovery, reconfiguration or diversity. Detecting and handling failures of the enclosing system is the responsibility of the user of the certified libraries.

39.3.7. 17

Covered by the Certified standard library API docs and the Rust type system, see 7.3.4.18-19 of EN 50716.

39.3.8. 18

N/A; no modelling is used, see Table A.17.

39.3.9. 19

Covered by the modular structure of the certified libraries:

The certified libraries have no overall architecture and a rather simple high-level design. Each module provides functions and data structures around a single well-defined topic. All modules have a doc-comment describing the design and contents of the module.

The selected set is 1, 5, 15, 17 and 19, which is an approved combination for SIL 2 (requirement 2 of this table).

39.4. Table A.4 — Software design and implementation (7.3, 7.4 and 7.5)

39.4.1. 1

N/A; no modelling is used, see Table A.17.

39.4.2. 2-5

Covered by

  • Architecture and software design (modular decomposition into modules with doc-comments)

  • Table A.20 below, for measure 4

  • the coding standard, see Table A.12 below, for measure 5

39.4.3. 6

Covered by Programming language and Code coverage report.

39.4.4. 7

Covered by Programming language.

The selected set is 2, 3, 4, 5, 7 and 8, which is an approved combination for SIL 2 (requirement 2 of this table).

39.4.5. 8

Covered by Programming language.

The selected set is 2, 3, 4, 5, 7 and 8, which is an approved combination for SIL 2 (requirement 2 of this table).

39.5. Table A.5 — Software component analysis and testing (6.2 and 7.4)

39.5.1. 1

Not used; no formal proof is performed. The static guarantees of the Rust language (type checking, borrow checking) provide an equivalent assurance, see Programming language.

39.5.2. 2

Covered by rustc lints and the tidy test suite, see Table A.19.

39.5.3. 3

The certified libraries are tested as laid out in the Testing Plan, and those test results, for all qualified targets, are available in the Test results overview.

The selected techniques of Table A.13 are documented in that table.

39.5.4. 4

Not used; no further metrics are measured. Size and complexity of the source code are limited by the language and the tooling, see Recursion limit and rustc lints.

39.5.5. 5

Covered by Code coverage report, see Table A.21.

39.5.6. 6

N/A; timing performance depends on the system requirements, which are unknown during the certification phase.

39.5.7. 7

Covered by Unit tests: every public function is tested through its interface, including its panic and error paths.

The selected set is 2, 3 and 5, which is an approved combination for SIL 1 to SIL 4 (requirements 1 and 2 of this table).

39.6. Table A.6 — Software integration analysis and testing (7.3 and 7.6)

39.6.1. 1

Covered by Testing Plan (the complete test suite is executed for every change on all qualified targets).

39.6.2. 2

N/A; timing performance depends on the system requirements, which are unknown during the certification phase.

39.7. Table A.7 — Overall software analysis and testing (6.2 and 7.2)

39.7.1. 1

N/A; timing performance depends on the system requirements, which are unknown during the certification phase.

39.7.2. 2

The certified libraries are tested as laid out in the Testing Plan, and those test results, for all qualified targets, are available in the Test results overview.

The “overall software” of a library is its complete public API; the black-box tests are the doc-tests and the integration tests of the Ferrocene components that use the certified libraries.

39.7.3. 3

N/A; no modelling is used, see Table A.17.

39.8. Table A.8 — Intentionally left blank

Intentionally left blank in the standard.

39.9. Table A.9 — Software quality assurance (6.5)

39.9.1. 1

Ferrous Systems is ISO 9001:2015 certified, see Qualification Plan - Ferrocene Organization and Internal procedures.

39.9.2. 2

Not used as method 1 is already applied.

39.9.3. 3

Ferrous Systems is ISO 9001:2015 certified, see Qualification Plan - Ferrocene Organization and Internal procedures.

39.9.4. 4

Covered by Build and Testing Process, Change Tracking and Configuration.

39.9.5. 5

Covered by the merge requirements of Development Process and the review checklists of Patching process.

39.9.6. 6

Covered by Requirements traceability report and Tracing requirements to tests.

39.9.7. 7

Covered by Validation Strategy: the CI infrastructure keeps logs and artifacts of every activity, and the Known Problems database records field data, see Known Problems Tracking.

39.10. Table A.10 — Software maintenance (9.2)

39.10.1. 1

Deviation: no per-change impact analysis is performed. Every change triggers a complete rebuild and a full execution of all test suites of the certified libraries, therefore the impact of a change is always the whole certified subset, see Re-verification scope.

39.10.2. 2

Covered by Known Problems Tracking and Change Tracking.

39.11. Table A.11 — Data preparation techniques (8.4)

N/A; the certified libraries are not configurable by application data.

The certified libraries are delivered as pre-compiled libraries and have to be statically linked into applications (see LIBS_CSTR_0080_LINK_STATIC). There is no way to configure a certified library binary after it is compiled and linked statically.

39.12. Table A.12 — Coding standards

39.12.1. 1

The Rust project has extensive measures (lints and tests) in place to assure quality and consistency of the codebase. The certified libraries use the same implicit standards as are ensured in the upstream codebase, to minimize divergence. Increased divergence from upstream leads to a higher maintenance burden and is a source of potential bugs.

As such, the certified libraries do not have an additional coding standard of their own. The upstream style and consistency rules are enforced by the tidy test suite (rustfmt, clippy, rustc lints), see Tidy test suite and rustc lints.

39.12.2. 2

The Rust project has extensive measures (lints and tests) in place to assure quality and consistency of the codebase. The certified libraries use the same implicit standards as are ensured in the upstream codebase, to minimize divergence. Increased divergence from upstream leads to a higher maintenance burden and is a source of potential bugs.

As such, the certified libraries do not have an additional coding standard of their own. The upstream style and consistency rules are enforced by the tidy test suite (rustfmt, clippy, rustc lints), see Tidy test suite and rustc lints.

39.12.3. 3

Covered by Programming language (compiler recursion limit, functions with many parameters are discouraged). Requirement 1 of this table is applied: further modularization of some complex generic functions is not sensible.

39.12.4. 4

Covered by the Rust language itself: a function has exactly one entry point and no goto jumps are available outside of inline assembly, see Rust.

39.12.5. 5

Covered by the Rust language itself: mutable global state requires unsafe code and explicit static mut or synchronization primitives; the certified subset avoids it, and the lints flag violations, see rustc lints.

39.13. Table A.13 — Dynamic analysis and testing

39.13.1. 1

Covered by Unit tests: the upstream test suites contain a large number of hand-written test cases, which are reviewed by upstream reviewers and by Ferrous Systems.

39.13.2. 2

Not used; no error seeding is performed.

39.13.3. 3

Covered by Code coverage report: test execution is measured against the code structure for every release.

39.13.4. 4

Not used; no cause consequence diagrams and no prototyping are used.

39.13.5. 5

Not used; no cause consequence diagrams and no prototyping are used.

39.13.6. 6

The test cases in unit test test suites are crafted with a lot of care. Ferrous Systems did not do a full review to ensure boundary and extreme values are always tested. But achieving 100% region coverage will ensure all code paths have been executed and no untested code exists due to no test with a specific input.

39.13.7. 7

Covered by Unit tests and the doc-tests: the equivalence classes of each function are documented in its doc-comment (including the panicking conditions) and tested.

39.13.8. 8

N/A; no process simulation is used. The tests run natively on the qualified targets or on emulators, see Infrastructure.

39.14. Table A.14 — Intentionally left blank

Intentionally left blank in the standard.

39.15. Table A.15 — Suitable Programming Languages

39.15.1. 1-7

Covered by

39.16. Table A.16 — Intentionally left blank

Intentionally left blank in the standard.

39.17. Table A.17 — Modelling

N/A; no modelling techniques are used. The design of each module is described in its doc-comment , see Architecture and software design. As there is no state managed or complex interdependencies of modules present inside of the library, no further modelling based approach or graphical representation is required to fully define the functions of the library.

39.18. Table A.18 — Performance testing

N/A; timing and memory performance depend on the target system and on the application, which are unknown during the certification phase (see 7.9.2.14 of IEC 61508-3).

39.19. Table A.19 — Static analysis

39.19.1. 1, 2

Covered by the Ferrocene compiler, which performs thorough control flow and data flow analysis (including the borrow checker), see Programming language.

39.19.2. 3

Covered by Failure analysis and Tool Analysis.

39.19.3. 4

Covered by Responsibility split: upstream changes are reviewed by appointed Rust experts who were not involved in the change, and Ferrous Systems reviews every pulled change.

39.20. Table A.20 — Components

39.20.1. 1

Covered by Programming language: all items of a module and all fields and methods of a data structure are private unless explicitly made public.

39.20.2. 2

Covered by Programming language (clippy::too_many_arguments).

39.20.3. 3

Covered by the Certified standard library API docs and the Rust type system.

39.21. Table A.21 — Test coverage for code

39.21.1. 1

Covered by Lifecycle Phases Overview (100 % statement coverage of the certified subset) and by Code coverage report.

39.21.2. 2-5

Not measured; the code coverage report measures region/statement coverage.

Requirement 4 of this table is applied: lines which cannot be covered by automated tests are annotated with #[ferrocene::annotation(REASON)] and reasoned there, see Manual test coverage. Additionally, unused code cannot be executed, see Uncertified and unused code.