Rendered from docs/obligations/0215-a-bless-run-of-the-conformance-render-tests-races-two-tests-over-one-fixture-file.md in the Headwater corpus. Every document on this half of the site is typed by the taxonomy the descriptor names: corpus.json.

A bless run of the conformance render tests races two tests over one fixture file

Context

The integrator of #959 in run 20260924-0411 found this race on main 0b6f8231. The product owner ruled it Record, because only this repository's test runs meet it.

Obligation

In engine/crates/conformance/tests/render.rs, under HEADWATER_BLESS=1, the_recorded_block_is_the_one_this_width_produces reads fixtures/wrapped.report. A sibling test in the same process rewrites that file. The reader panicked with "the recorded block has lines". The test passes without bless, and it passed under bless at 82fc576a.

Discharge

This record discharges when no two tests in one process read and write the same fixture under bless, and a bless run of the suite passes on repeated runs.

#1162 discharged this record on 2026-09-28. It removed the_recorded_block_is_the_one_this_width_produces, which was the reader. The writer, the_report_wraps_every_line_to_the_width, made the same two width assertions against the rendered string, with one difference. The reader did not count a line of one word toward the lower bound, and the writer did. A maximum over all lines is never smaller than a maximum over some of them, so the writer's bound was the weaker one. One word longer than the column could hide a fill that stops short on every other line. The same change added that filter to the writer. A scratch mutation shows the gap. It stopped the fill 20 columns short and put a 69-character URL into a remediation. The old bound passed that render, and the new bound failed it. Without bless, compare fails when WIDTH moves and the file does not. So wrapped.report now has one case that names it, and the file did not change. A local test put a 200 ms empty file into the blessed write. With that window, 9 of 20 blessed runs failed before the change. Each failure was the panic that this record names. After the change, 0 of 20 failed. Without the window, 0 of 50 blessed runs of the render target failed, and 0 of 20 blessed runs of the whole crate failed. This record repeats the render.rs half of HW-OBL-0176, and the same change discharges both.