DO-178C, IEC 61508 and ISO 26262 all require evidence that deadlines are met. The evidence is produced today by proprietary tools priced per seat, and reproducing their result means trusting their binary. The standard asks for proof and the industry supplies a receipt.
A certificate here is a file: the task set, the derivation step by step, the tool version, and a hash over all of it. An auditor runs the same tool on the same input and gets the same bytes, or the certificate is void.
Nothing is implemented. This repository exists so the design can be argued with before it is built, and so the name is not claimed by something that does less.
The hard part is not the signature. It is deciding what a derivation must contain to be checkable by somebody who was not there — and that is a question about what an auditor needs, not about cryptography.
dy-wcet is the piece that works
today: response-time analysis for fixed-priority task sets, in integer
arithmetic that refuses rather than rounds.
Apache-2.0 OR MIT, at your option.