31562104955
93%

Ran 12 Aug 2026 04:05AM UTC

Jobs 1

Files 1

Run time 1min

Badge

Embed ▾

Committed 12 Aug 2026 04:04AM UTC coverage: 94.681%. Remained the same

Build # 31562104955

Build Type

push

github

Committed by

web-flow

Commit Message

Calibrate hw-ct max-spread gate for the full-sign warmup transient (#43)

* Calibrate the hw-ct max-spread gate for the full-sign warmup transient

The 30 MHz max-spread determinism policy failed intermittently on the
fixedsign fixture. Per-sample logs show why: the ~117M-cycle sign's measured
trials split into two clusters ~1056 cycles apart, first-two-trials vs
last-two, and the split is *identical for both keys* (Welch t≈0). It's a
one-time warmup/settling transient, not a key-dependent path — the 32-cycle
absolute tolerance (0.27 ppm over 117M cycles) was simply too tight to
survive it, failing when the settle straddled the measured trials and passing
when all four landed settled.

Raise warmup_blocks 1→4 so the transient lands before the measured trials,
and set positive_max_spread to a realistic 4096-cycle bound for the
multi-million-cycle sign workloads. The between-key Welch gate (threshold 4.5)
is unchanged and remains the actual leak detector; determinism was never a
property of the production signer anyway, which is blinded.

Coverage Stats

979 of 1034 relevant lines covered (94.68%)

17235.38 hits per line

Jobs

ID	Job ID	Ran	Files	Coverage
1	31562104955.1	12 Aug 2026 04:05AM UTC	1	94.68	GitHub Action Run

kaidokert / krabiecdsa-rs / 31562104955
93%

README BADGES
x

Markdown

Textile

RDoc

HTML

Rst

Jobs

Source Files on build 31562104955

kaidokert / krabiecdsa-rs / 31562104955 93%

README BADGES x

Markdown

Textile

RDoc

HTML

Rst

Jobs

Source Files on build 31562104955

kaidokert / krabiecdsa-rs / 31562104955
93%

README BADGES
x