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

stillwater-sc / universal / 42529495092
89%
master: 84%

Build:
Build:
LAST BUILD BRANCH: fix/fixpnt-quire-resolve-saturates
DEFAULT BRANCH: master
Ran 08 Nov 2026 10:43PM UTC
Jobs 1
Files 871
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

08 Nov 2026 10:17PM UTC coverage: 88.867% (+0.03%) from 88.835%
42529495092

push

github

web-flow
fix(ereal): NaN comparisons, signed zero, numeric_limits floor, and efloat print/parse at the exponent extremes (#2020)

ereal:
- compare_adaptive read a NaN limb as equal to anything, so nan == 5,
  nan >= 5 and nan <= 5 were true. NaN is now unordered in every
  operator, against ereal and against double. The >= forms were
  !(a < b), which is true for a NaN.
- min and max follow C99 fmin and fmax: a NaN operand is ignored.
  max(1, NaN) was NaN, while max(NaN, 1) was 1. fmin and fmax are added.
- signbit(-0) was false, abs(-0) was -0, and to_string(-0) printed 0.
  All three now read the sign bit.
- numeric_limits min() is the full-precision floor of the type's own
  limb count, 2^(min_exponent(F) - 1 + digits(F) (N - 1)), as #1882
  chose for dd, qd and the cascades (user decision). It was the
  two-limb floor 2^-969 for every N, while min_exponent and minpos()
  said 2^-1022. min_exponent follows min(); has_denorm is present, and
  denorm_min() is minpos().
- parse saturates the decimal exponent. ereal("1e4294967297") wrapped
  the int and returned 1e1. isdigit takes the character as an
  unsigned char.
- Fixed-format to_string rounds a value below 10^-p into the last
  digit: 0.006 at two digits is 0.01, not 0.00. A rounding carry no
  longer drops a digit: 0.0996 at two digits is 0.10, not 0.1.

efloat:
- to_string held the decimal exponent in an int, which efloat's binary
  exponent overflows. It also scaled by 10 one step at a time, which
  hung on max() and lost digits at 1e5000. The exponent is now 64-bit,
  and the value is scaled once by a power of ten built by squaring.
  1.234e5000 prints in 1.9 us instead of 140 us.
- parse read the exponent as an int32 and refused literals beyond about
  +-13000, leaving NaN even for 0e100000. It now reads a saturating
  64-bit exponent and returns a signed zero for a zero mantissa.
  Literals the exact converter cannot reach are scaled in efloat
  arithmetic, faithfully rounded. Beyond the type's ran... (continued)

149 of 157 new or added lines in 5 files covered. (94.9%)

6 existing lines in 3 files now uncovered.

60132 of 67665 relevant lines covered (88.87%)

9673792.85 hits per line

Uncovered Changes

Lines Coverage ∆ File
8
95.53
0.04% include/sw/universal/number/efloat/efloat_impl.hpp

Coverage Regressions

Lines Coverage ∆ File
3
95.53
0.04% include/sw/universal/number/efloat/efloat_impl.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
Jobs
ID Job ID Ran Files Coverage
1 42529495092.1 08 Nov 2026 10:43PM UTC 871
88.87
GitHub Action Run
Source Files on build 42529495092
  • Tree
  • List 871
  • Changed 7
  • Source Changed 5
  • Coverage Changed 6
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #42529495092
  • 45c4de13 on github
  • Prev Build on main (#42523560319)
  • Next Build on main (#42539818393)
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