1. What cost the most. The two waits for builds: the first ran past the 120 s foreground default, so I ran it a second time in the foreground. Then the gate, which was short. Writing proved.nika took three tries. I guessed str.parse_i64(), cli::args().collect() and a map over a BTreeMap iterator, and none of them exist in that form. Then the header’s indexes, carries a comma, which gave a total of 0.
2. What the brief could have handed me.
   • The text parse is text::parse_i64(s) in std::text.
   • Scripts collect their arguments with for a in cli::args() { args.push(a) }.
   • A text may only be cut at a position find returned (NK1214).
   • --bounds and --optimization are already global flags, so no new flag on LowerStd was needed. Reading clap to find that out took a turn.
   • The exact header format of bounds_report.
   • The lowering checks with the full beside parses; without_bodies is only used for the cache key.
3. Checks run: cargo check, then one build, two lower-std --bounds runs, the script three times, the --lib sysroot:: bounds_report:: unit tests and the gate (green). Wasted: the second build call. The first one was still running in the background, so the second only waited on it. Also an awk probe for line 63, which found nothing because the site is on line 64 now.
4. The change that would save the most: a “Nikaia forms for scripts” cheat sheet in builder-facts. It would cover text to integer, collecting args, iterating a map, sorting tuples, and how to cut text. Second: when the brief knows a commit closes only one step of an issue, it should say so. Then the Fixes/Part line comes out right the first time, without amending the message.

This site is open source. Improve this page.