tru builds and tests clean against its siblings' origin default branches

date: 2026-09-23 · property #39 · repo: tru

claim

property 39 asks whether every phase-1 component builds and tests from a clean checkout of its default branch against the default branches of its siblings, with no dead path, no version pin behind a sibling, and no crate present only in an owner's working tree. the sweep of 2026-09-22/23 found bbg, cybergraph, foculus, mudra, neuron, nox, prysm, zheng, true-cyber, glia, honeycrisp, rune, vault and radio failing this gate. tru was not in that list. this measures it directly.

method

origin/master of tru checked out in an isolated worktree (89416291dd4cc993a79d63cbd3e2445294b173d3). tru's three path dependencies are strata's nebu and core/compute/ext crates, hemera, and honeycrisp's acpu/unimem. strata's owner working tree already matched its origin default branch (56aedb2d12b3126c601eb333419136d403614dbb); hemera and honeycrisp did not (hemera's tree carried unpushed local commits, honeycrisp's tree was one commit behind origin and several ahead), so each was checked out fresh from origin/main into a detached worktree and substituted for the owner's tree for this run: hemera at 23f3bbcff910ea6d504ceb505680a539260869da, honeycrisp at 9593c7218e14e2b5b816d9008b45ab60b9580cf3.

result

$ cargo check --tests --workspace
    Checking cyber-tru v0.1.1 (rs)
    Checking tru-cli v0.1.0 (cli)
warning: unused import: `super::*` (rs/pass/embed.rs:173, pre-existing, unrelated)
    Finished `dev` profile [unoptimized + debuginfo] target(s) in 1.05s

$ cargo test --workspace
test result: ok. 85 passed; 0 failed; 0 ignored (lib)
test result: ok. 1 passed; 0 failed; 0 ignored (tests/honesty_loop.rs)
test result: ok. 1 passed; 0 failed; 0 ignored (tests/smoke.rs)
test result: ok. 0 passed; 0 failed; 0 ignored (cli)
test result: ok. 0 passed; 0 failed; 1 ignored (doc-tests)

87 tests pass, 0 fail, on rustc 1.98.0 (88d9e12ae 2026-08-18). no dead path dependency, no pin behind a sibling's published version, no crate resolved only from an owner's working tree. tru closes its row of property 39; the row stays open until every listed repo is covered.

caveat

this is one repo's slice of a multi-repo row. it says nothing about tru's own dependents (plumb/tok, foculus) resolving tru correctly, and nothing about the fourteen repos the sweep already found broken.

re-run 2026-10-17

the 2026-09-23 numbers above are kept as measured. re-run during the 2026-10-17 branch triage, on origin/master of tru at 74ecfb13517fad364048add5308e7c7feaa40d19 plus this page, in a detached worktree, against detached worktrees of the siblings' origin/main: strata 494df7a, hemera 1c385e5 (cyber-hemera 0.3.1), and honeycrisp 3ffc0d5 (its main checkout was clean and equal to origin, so it was linked rather than copied). rustc/cargo 1.95.0 (Homebrew).

$ cargo fmt --all --check                                         # clean
$ cargo clippy --locked --workspace --all-targets -- -D warnings  # clean
$ cargo test --locked --workspace
test result: ok. 164 passed; 0 failed; 0 ignored (lib)
test result: ok. 1 passed; 0 failed; 0 ignored (tests/honesty_loop.rs)
test result: ok. 1 passed; 0 failed; 0 ignored (tests/smoke.rs)
test result: ok. 0 passed; 0 failed; 0 ignored (cli)
test result: ok. 0 passed; 0 failed; 1 ignored (doc-tests)
$ cargo build --locked --workspace --examples                    # 0 warnings

166 tests pass, 0 fail, 0 warnings. the same day the GitHub CI went green for the first time: it had failed on every run because the workflow did not check out honeycrisp, whose acpu manifest cargo reads even on linux (tru#38). tru still holds its slice of property 39.