docs(agents): correct the Python measurement; the first figures were wrong

sean commented

2026-09-10 14:04:47 +00:00

Owner

The note claimed one test spent about 59 seconds of interpreter start-up inside a 112-second test. Those numbers were taken while a full guard suite and other jobs ran on the same machine. Re-measured quiet, that test takes 13.0 seconds, not 112, and the start-up share was over-attributed. What is true, measured quiet: binjovi's make contract-check took 24.4 seconds with six Python steps and 3.6 seconds without them, so the interpreter was about 85 percent of that target's run time. The 2,806 fixture invocations stand, counted by instrumenting the fixture, and so does python3 at 18-21 ms against jq at 2.7 ms over 200 iterations. The conclusion is unchanged and now rests on figures that hold up. https://claude.ai/code/session_01KZoQin34jeyt6nDGqvJA76

sean added 1 commit

2026-09-10 14:04:47 +00:00

docs(agents): correct the Python measurement; the first figures were wrong

binjovi/ci Binjovi completed the frozen plan

Details

83fb01e3ed

WHAT WAS WRONG. The note said one test file spent "about 59 seconds of
interpreter start-up inside a 112-second test", and that this was most of 69
seconds of CPU. Those numbers were taken while a full guard suite and several
other jobs were running on the same machine. Re-measured quiet, that test takes
13.0 seconds, not 112, and the start-up share was over-attributed. A
justification with a wrong number in it is worse than no number.

WHAT IS TRUE, MEASURED QUIET. binjovi's `make contract-check` took 24.4 seconds
with six Python steps in it and 3.6 seconds without them, so the interpreter was
about 85 percent of that target's run time. The fixture invocation count stands:
2,806 starts in a single run, counted by instrumenting the fixture. So does the
start-up comparison: python3 in 18 ms, 21 ms once it imports json, os, sys and
urllib, against 2.7 ms for jq, over 200 iterations each.

The conclusion is unchanged and now rests on figures that hold up.

Claude-Session: https://claude.ai/code/session_01KZoQin34jeyt6nDGqvJA76

binjovi-bot merged commit 83fb01e3ed into trunk

2026-09-10 14:11:47 +00:00

Rows
Columns

docs(agents): correct the Python measurement; the first figures were wrong #572