stillwater-sc / universal / 23224573073 / 1
85%
master: 84%

LAST BUILD BRANCH: feat/issue-1061-online-muldiv
DEFAULT BRANCH: master
Ran
Files 627
Run time 15s
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
coverage: 84.312% (+0.007%) from 84.305%
23224573073.1

push

github

web-flow 
feat(cmake): enable serialization tests in CI_LITE (#590)

* feat(cmake): enable serialization tests in CI_LITE

Add UNIVERSAL_BUILD_LINEAR_ALGEBRA_DATA to the CI_LITE cascade so
serialization tests (data_serialization, data_typed_serialization,
data_translation) are built and run in CI on every PR.

Previously these tests only ran with UNIVERSAL_BUILD_ALL, which is
not used in CI. This meant serialization regressions (like the
lns::fraction() stub in #509) were invisible to CI.

Resolves #586

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix(blas): add value verification to serialization round-trip tests

Tests 6-8 (lns, dbns, integer) now verify bit-exact fidelity by
comparing the serialized text after a save->restore->re-save cycle.
If the re-serialized text differs from the original, the round-trip
lost data.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix(blas): guard value comparison on successful restore

Make the re-save/compare check conditional on restore succeeding
so a failed restore doesn't double-count as a value mismatch.
Applied to lns, dbns, and integer round-trip tests.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix(fixpnt): silence uninitialized warning in numeric_limits

fixpnt's epsilon(), round_error(), and denorm_min() declared a local
FixedPoint without initialization, then called setbit() on it. GCC
warned that the underlying blockbinary storage was uninitialized.

Fix: value-initialize with {} and call clear() before setbit() to
ensure all bits are zero before setting the ULP bit.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix(blockbinary): silence GCC -Warray-bounds false positive in setbit

GCC at -O3 produces false-positive -Warray-bounds warnings when
template inlining causes it to confuse blockbinary instantiations
across fixpnt's convert() -> setbit() chain. The existing guard
(blockIndex < nrBlocks) is correct but GCC's analyzer loses track
of which instan... (continued)

44203 of 52428 relevant lines covered (84.31%)

5934176.39 hits per line

Source Files on job 23224573073.1