| name | orlix-upstream-conformance |
| description | Use when building, running, triaging, or claiming success for upstream Linux, upstream mlibc, Coreutils, or other upstream project test suites in Orlix. Requires pristine upstream sources and upstream tests; failures must be routed to the owning Orlix layer instead of editing generated upstream trees or adapted tests. |
Orlix Upstream Conformance
Upstream conformance work proves that Orlix matches upstream-owned behavior. The upstream source and upstream tests are authoritative.
Rules
- Use pristine upstream sources and tests.
- Do not edit, filter, adapt, or reinterpret upstream tests as success.
- Do not modify upstream tests through durable patch stacks used by conformance schemes. Extra regressions belong in Orlix-owned tests, not in patched upstream test trees.
- Do not patch generated upstream trees such as
Build/OrlixMLibC, Build/OrlixOS, or Build/OrlixKernel/src/linux-*-port. Read them freely for diagnosis; route durable changes to owning Orlix inputs.
- Compare failures against upstream Linux, upstream mlibc, or the relevant package behavior before changing Orlix.
- Treat every durable upstream patch and package feature override as suspect until root cause is proven. Before adding or changing one, prove the lower owning layers are correct: Linux/kernel syscall and filesystem behavior, OrlixOS rootfs/package inputs, and reviewed package toolchain configuration.
- Route fixes by ownership: kernel semantics to
OrlixKernel, libc behavior to OrlixMLibC/Sources, upstream package/rootfs assembly and delivered OS payload/session wiring to OrlixOS, and private iOS mechanics to OrlixHostAdapter.
- Test runners may consume the
OrlixOS session surface to collect upstream output, but they must not patch, filter, skip, or reinterpret upstream tests.
- Do not disable upstream package features, suppress upstream tests, or force
configure/libtool with ad hoc LD overrides to pass cross builds; fix reviewed OrlixOS package-toolchain inputs and prove the unchanged upstream behavior.
Evidence
Record exact commands, upstream versions and commits, logs, failures, skips, crash checks, final markers, and the patch or override root-cause chain in the workflow's structured report under Build/AgentHarness/.
For Coreutils full-suite claims, success requires the upstream suite marker with zero failures and zero skips for the expected test count. Partial or targeted runs are not full proof.
