Re: Reporting script runtimes in pg_regress

Re: Tom Lane 2019-02-18 <28360(dot)1550506699(at)sss(dot)pgh(dot)pa(dot)us>
> >>> We should also strive to align "FAILED" properly.
>
> >> Yeah, not strictly required, but someone might want to play around with
> >> it a bit.
>
> > FWIW I don't think we localize pg_regress output currently, so that
> > argument seems moot ... But I think we can get away with constant four
> > spaces for now.
>
> I pushed Peter's suggestion for %8.0f; let's live with that for a little
> and see if it's still annoying.

The ryu changes make postgresql-unit fail quite loudly:

$ cat regression.out
test extension ... ok 359 ms
test tables ... FAILED 164 ms
test unit ... FAILED 867 ms
test binary ... ok 20 ms
test unicode ... ok 18 ms
test prefix ... FAILED 43 ms
test units ... FAILED 207 ms
test time ... FAILED 99 ms
test temperature ... FAILED 22 ms
...

The misalignment annoyed me enough (especially the false alignment
between "ms" on the first row and "164" on the next row) to look into
it. Aligned it looks like this:

test extension ... ok 399 ms
test tables ... FAILED 190 ms
test unit ... FAILED 569 ms
test binary ... ok 14 ms
test unicode ... ok 15 ms
test prefix ... FAILED 44 ms
test units ... FAILED 208 ms
test time ... FAILED 99 ms
test temperature ... FAILED 21 ms
...

It doesn't break translations because it prints the extra spaces
separately.

In run_single_test() (which this output is from), it simply aligns the
output with FAILED. In run_schedule(), there is the 3rd output string
"failed (ignored)" which is considerably longer. I aligned the output
with that, but also made the timestamp field shorter so it's not too
much to the right.

Christoph