1. What cost the most: the full tests/compiler --both-settings run, about 10+ minutes of waiting. The edit itself was quick because the brief named both places to change. I also spent two turns on row b. The brief said it prints 1 today, but it failed in rustc on the suite binary. I re-probed it in a second shape to rule out a mistake in my own probe.
2. What I looked up:
   • nikaia test --help, to find out whether it filters by name. It doesn’t, yet the brief says to run tests “by name”.
   • That views_the_call_s has a third caller, names_borrowing in views.nika:463. The brief named only two.
   • The emit_boxed_future_in helper and how the crossing suffix is written in emit_core_types.

   None of these was expensive.

3. Checks run:
   • Rows a, b, c and the Vec variant before and after the fix, on both binaries.
   • The cycle (two lower-std runs and the build), plus a second build to confirm the fixpoint.
   • cargo fmt --check, clippy at 1.97, clippy at 1.99.1 on nikaia-std with --features compiler, pre-push-check.sh.
   • All of tests/compiler.

   Partly wasted: clippy at 1.99.1, because only lowered files changed and their lints are allowed. The full tests/compiler run was unavoidable without a name filter, and it ran about 472 tests to check about 45.

4. One change: in the brief template, write “the tests are run by running all of tests/compiler once” instead of “run by name”. Or better, add a name filter to nikaia test (--filter SUBSTR), which would cut most of the 10-minute wait for a brief like this. Also, check the claimed “today” result of each row before filing it: row b was stated wrongly.

This site is open source. Improve this page.