1. Most time: the brief’s test run, about 6 min of it plus the 7 comptime_* programs. The one rework was clippy’s while_let_loop on the lowered run_fingerprint loop, which cost a second clippy run. The run took about 25 turns in all.
2. Looked up, could have been handed:
   • Run 2’s commit adds no row to scripts/self_hosting.py. I had to grep for that, then find the right spot.
   • The format of the plan row’s “built” cell. I copied K3a’s.
   • The tools module’s allow list in lib.rs.
   • Whether 1a5fdf6d is on HEAD. --grep='#1106' found the architect’s batch commit, not the fix, so I checked the hash.
3. Checks run:
   • The two lower-std runs and the build, plus a third pass for the fixpoint (it changed nothing).
   • The it filters and comptime_* in a scratch project.
   • --lib comptime_run::: wasted, the module has 0 tests.
   • The run_* tests: there are none in tests/compiler.
   • clippy on 1.97, then on 1.99.1 in a separate target dir, about 10 min. That one only confirmed green, but the rules require it.
   • fmt and pre-push.
4. One change: a run-N brief that cherry-picks a wip commit should say which bookkeeping that commit lacks (the self_hosting.py row, the plan row, a lint the lowered file is known to trip). It should also drop test names that match nothing (run_*, --lib comptime_run::).

This site is open source. Improve this page.