1. What cost the most: the pre-flight. The hashes 788981d and 76c216e had been rebased, so the brief’s git log | grep found nothing. It took 3 turns to match them by subject (6816f47, ce758cb). After that, the scratch sysroot for the assert check took 3 tries. I didn’t know its layout (<root>/nikaia-std/src/tools, plus described.contracts), and I wrote assert x where the syntax is assert(x).
2. What I looked up that the brief could have handed me:
   • the asserts in header format (check_facts.nika :805);
   • that check_against always refuses unproved claims, so refuses is true and an unproved assert gets the whole file refused;
   • the text-method names Nikaia has (matches, split_once);
   • the sysroot layout that tools_dir expects.

   The brief’s “13 panics in 9 files” had no counting rule behind it, so I had to work out why I got 3 in 3.

3. Checks run:
   • the build, once, plus nikaia check on the script;
   • lower-std --bounds --asserts at both settings, and --bounds alone;
   • the scratch-sysroot claim checks and a synthetic header line;
   • the named unit tests, fmt, clippy 1.97 and 1.99.1, pre-push-check.sh.

   Wasted: the first lower-std --bounds run without --asserts, which a synthetic report could have covered. Also two sysroot runs that failed on layout and syntax. The git rev-parse probe was not wasted: it found the temporary-view defect.

4. One change: have the pre-flight match commits by subject (git log --grep), not by short hash, since the driver’s rebases change the hashes. Second choice: a figure the brief quotes, such as the panic count, should come with the command that produced it.

This site is open source. Improve this page.