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

stillwater-sc / universal / 41519197556
89%
master: 84%

Build:
Build:
LAST BUILD BRANCH: fix/fixpnt-quire-resolve-saturates
DEFAULT BRANCH: master
Ran 02 Nov 2026 08:51PM UTC
Jobs 1
Files 773
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

02 Nov 2026 08:26PM UTC coverage: 88.491% (+0.009%) from 88.482%
41519197556

push

github

web-flow
fix(cfloat): NaN, signed zero and precision in min/max and nextafter (#1902)

min/max were cfloat(std::min(double(x), double(y))): NaN handling depended on argument order, a signalling NaN came back signalling, signed zeros depended on argument order, and every cfloat wider than double was rounded. They now compare cfloat operands directly as IEEE-754-2019 minimum/maximum (a NaN in either argument gives a quiet NaN, -0 < +0). New fmin/fmax follow C fmin/fmax (C23 fminimum_num): a NaN operand, quiet or signalling, is ignored; two NaNs give a quiet NaN.

nextafter/nexttoward return a quiet NaN for a NaN operand (a NaN target used to be a direction, a NaN x had its payload stepped) and set errno = ERANGE when a finite x steps to infinity. nexttoward steps x by its own ulp and compares x and the cfloat<128,15> target in a format that holds both exactly (it used to step a 128-bit copy of x and round back, which never moved a narrower x).

operator++/-- step -0 like +0: ++(-0) wrapped to the qNaN pattern, so nextafter(-0, 1) returned NaN, and --(-0) without subnormals landed on a subnormal encoding. Without subnormals and with more than 64 fraction bits, ++(+0) now reaches minpos (setfraction only wrote 64 bits). The cfloat-to-cfloat converting constructor keeps the sign of zero.

Tests: math/next.cpp special cases across 11 configurations (half, single, quad, multi-block, max-exponent, no-subnormal up to 128 bits, cfloat<256,19>); new math/minmax.cpp with named cases and exhaustive pairs of three 8-bit configurations.

Resolves #1901

BEHAVIOUR-CHANGE: cfloat min/max return a quiet NaN for a NaN in either argument and order -0 below +0; nextafter/nexttoward return a quiet NaN for a NaN argument, nextafter(-0, 1) and ++(-0) give +minpos, and nexttoward steps by x's own ulp
BEHAVIOUR-CHANGE: converting a cfloat -0 to another cfloat configuration now gives -0 instead of +0

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

44 of 44 new or added lines in 3 files covered. (100.0%)

1 existing line in 1 file now uncovered.

56127 of 63427 relevant lines covered (88.49%)

8952035.72 hits per line

Coverage Regressions

Lines Coverage ∆ File
1
97.53
-0.05% include/sw/universal/number/cfloat/cfloat_impl.hpp
Jobs
ID Job ID Ran Files Coverage
1 41519197556.1 02 Nov 2026 08:51PM UTC 773
88.49
GitHub Action Run
Source Files on build 41519197556
  • Tree
  • List 773
  • Changed 3
  • Source Changed 3
  • Coverage Changed 3
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #41519197556
  • 089263ef on github
  • Prev Build on main (#41478103696)
  • Next Build on main (#41557127986)
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