• Home
  • Features
  • Pricing
  • Docs
  • Announcements
  • Sign In

stillwater-sc / universal / 43503536557
89%
master: 84%

Build:
Build:
LAST BUILD BRANCH: fix/fixpnt-quire-resolve-saturates
DEFAULT BRANCH: master
Ran 13 Nov 2026 09:05PM UTC
Jobs 1
Files 888
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst

13 Nov 2026 08:39PM UTC coverage: 89.05% (+0.04%) from 89.011%
43503536557

push

github

web-flow
fix(numeric): exact ereal remainder, elreal past double range, sorn saturation, floatcascade division, expansion_ops sums and native quire assignment (#2056)

* fix(numeric): exact ereal remainder, elreal past double range, sorn saturation, floatcascade division, expansion_ops sums and native quire assignment

The multi-component, elastic and sorn findings of QA sweep #2029.

sorn
- adaptToDT() clamped only the bound that crossed an end of a lattice without
  infinity, so 3.75 + 3.75 on sorn<0,4,8,..,no inf> was the inverted interval (7,4]
  and 3.75 * 3 was (8.75,4]. A lower bound at or past the top, or an upper bound at
  or below the bottom, now saturates to the outermost interval, as conversion does.

ereal
- fmod/remainder: exact_remainder subtracted Real(q) * b, a product with twice b's
  limbs cut to the limb budget, so ereal<4> fmod(1e30/7, 1/3) was off by ~1e-101
  relative to y. The reduction now runs on the raw expansions (expansion_product,
  expansion_sum_normalized) and is converted once: the result is exact whenever the
  exact remainder fits the limb budget. remainder's final r - |y| is also taken on
  the raw expansion, where it is exact and small though r itself may not fit.
- <= and >= build one temporary for a double operand, and do one comparison
  (they took < then ==, with a temporary for each).

elreal
- approx<T>() converted each block on its own: past T's range blocks saturated to
  +-inf and opposite-signed infinities summed to NaN (double(exp(elreal(1e6))) was
  NaN). The blocks are summed relative to the leading block's exponent and scaled
  once (approx_scaled), so a value past the range is +-inf; significand() scales by
  -scale first instead of dividing an inf.

floatcascade
- operator/= took two quotient digits, so floatcascade<3> and <4> divided at
  double-double accuracy. It is now long division with N + 1 digits, as qd_cascade
  does, closed by renorm5/renorm4.
- x / +0 was -inf (sign() of a zero is 0): the sign co... (continued)

383 of 413 new or added lines in 11 files covered. (92.74%)

7 existing lines in 4 files now uncovered.

62408 of 70082 relevant lines covered (89.05%)

11059460.31 hits per line

Uncovered Changes

Lines Coverage ∆ File
21
81.9
internal/floatcascade/arithmetic/division.cpp
6
93.95
-1.43% internal/expansion/api/expansion_ops.cpp
2
98.25
2.28% include/sw/universal/internal/expansion/expansion_ops.hpp
1
97.47
-0.57% include/sw/universal/number/ereal/math/functions/fractional.hpp

Coverage Regressions

Lines Coverage ∆ File
3
86.07
1.64% include/sw/universal/internal/floatcascade/floatcascade.hpp
2
34.7
-0.63% include/sw/universal/verification/test_suite_randoms.hpp
1
97.8
-0.06% include/sw/universal/number/cfloat/cfloat_impl.hpp
1
93.95
-1.43% internal/expansion/api/expansion_ops.cpp
Jobs
ID Job ID Ran Files Coverage
1 43503536557.1 13 Nov 2026 09:05PM UTC 888
89.05
GitHub Action Run
Source Files on build 43503536557
  • Tree
  • List 888
  • Changed 12
  • Source Changed 10
  • Coverage Changed 12
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #43503536557
  • c8e8d9b3 on github
  • Prev Build on main (#43483250202)
  • Next Build on main (#43514747065)
STATUS · Troubleshooting · Open an Issue · Sales · Support · CAREERS · ENTERPRISE · START FREE TRIAL · SCHEDULE DEMO
ANNOUNCEMENTS · TWITTER · TOS & SLA · Supported CI Services · What's a CI service? · Automated Testing

© 2026 Coveralls, Inc