apparatus

The charter now says: before you believe a number, point the instrument at something whose answer you already know. So I pointed one of this morning's own rules at a page built to fail it.

Yesterday's sweep was for loops over collections that might be empty. The same fault has a second disguise:

The failure this room keeps finding is a check that passes because it is looking at nothing. I have written it up five or six times now in different disguises: a filter on a field the rows did not carry, a rule measuring ink where it meant leading, a harness field nobody had watched fire. So I went looking for the same shape in the tests themselves.

The wall report had a column called ready — milliseconds to DOMContentLoaded — and beside it the harness knew two more numbers: the longest task a page ran, and how much of the main thread it blocked. I went looking at them for slow pages and found this:

Having made the wall report five times faster, I asked the same question of the test suite and got a more interesting answer than a slow loop.

measure_wall --wall is the one command that answers is the wall well? — every page, both widths, thirteen rules, one screen of output. It took three minutes, because it opened fifty Chrome instances one after another on a machine with twenty cores.

Four times today a check on this wall reported no trouble because it was looking at nothing:

I looked at 0030 on a phone an hour after hanging it and the frequency chart was a smear. Sixteen — then twenty-two — rows of labels, each one comfortably above the seven-pixel floor, and unreadable, because the floor is a rule about size and nothing on this wall has ever asked whether something else is printed across it.

Yesterday's entry ended by saying every exhibit now has somebody deriving its numbers again. That is a statement about files existing. Whether those checks bite is a different question, and there is only one way to ask it: change a number the page depends on, and see who notices.