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

stillwater-sc / universal / 42435409400
89%
master: 84%

Build:
Build:
LAST BUILD BRANCH: fix/fixpnt-quire-resolve-saturates
DEFAULT BRANCH: master
Ran 08 Nov 2026 01:55AM UTC
Jobs 1
Files 868
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 01:42AM UTC coverage: 88.771% (+0.002%) from 88.769%
42435409400

push

github

web-flow
fix(einteger): negative shift direction, +0 after >>=, round-once float conversion, any native operand (#1994)

- operator>>= with a negative count called itself with -shift, so 8 >>= -2
  gave 2. It now shifts left, as <<= with a negative count already shifted
  right.
- A negative value shifted to zero kept its sign (isneg() and iszero() both
  true). >>= now leaves +0. It still truncates the magnitude toward zero, as
  a sign-magnitude type does (-5 >> 1 is -2); documented.
- Conversion to float/double/long double added 2^i per set bit, rounding at
  every add past 2^digits: (float)(2^30 + 2^29 + 2^6 + 1) came out 128 too
  low. It now gathers the top digits bits exactly, rounds once to nearest
  even with a round and a sticky bit, and scales with ldexp; a value beyond
  the range becomes infinity.
- The native-operand overloads were long long for + - * / % plus a stray
  unsigned long long for /, so a / 2 was ambiguous and did not compile. One
  constrained template per operator and side now takes any native arithmetic
  type; integers convert exactly, floating-point truncates as before.

Tests: api/shift.cpp covers negative counts and the sign of zero;
conversion/assignment.cpp checks float/double rounding against native
conversion, including ties and values wider than 64 bits; the new
arithmetic/scalar_operands.cpp checks every native operand type on both
sides. All three fail (or do not compile) against the previous header.

Resolves #1987

BEHAVIOUR-CHANGE: einteger a >>= -k now shifts left by k (was right by k)
BEHAVIOUR-CHANGE: einteger conversion to float/double/long double is now correctly rounded; values above 2^24 (float) or 2^53 (double) can differ in the last bit from before

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

27 of 27 new or added lines in 1 file covered. (100.0%)

2 existing lines in 1 file now uncovered.

59601 of 67140 relevant lines covered (88.77%)

9637623.34 hits per line

Coverage Regressions

Lines Coverage ∆ File
2
91.45
-1.71% include/sw/universal/verification/correctly_rounded_reference.hpp
Jobs
ID Job ID Ran Files Coverage
1 42435409400.1 08 Nov 2026 01:55AM UTC 868
88.77
GitHub Action Run
Source Files on build 42435409400
  • Tree
  • List 868
  • Changed 3
  • Source Changed 1
  • Coverage Changed 3
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #42435409400
  • 218f4d95 on github
  • Prev Build on main (#42433863613)
  • Next Build on main (#42446138168)
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