Debian Patches

Status for yade/2026.1.0-8

Patch Description Author Forwarded Bugs Origin Last update
10_disable_some_tests.patch Disable some patches
===================================================================		 Anton Gladky <gladk@debian.org> no 2021-11-09
remove-python3-ipython-genutils Remove python3 ipython genutils=================================================================== no
1156.patch [PATCH 1/2] Boost 1.88 fix #401: when using ThinRealWrapper<long double> disable std::fpclassify and constexpr detection Janek Kozicki <jkozicki-yade@pg.edu.pl> no 2026-01-23
fix-checkShapeVolume-numpy-import.patch Fix NameError: name 'np' is not defined in checkShapeVolume.py Anton Gladky <gladk@debian.org> no debian 2026-06-22
fix-viscoelasticpm-maybe-uninitialized.patch Avoid GCC -Wmaybe-uninitialized false positive in Ip2_ViscElMat_ViscElMat_ViscElPhys GCC's -Wmaybe-uninitialized misfires on the chained Eigen expression
templates (auto holding a lazy CwiseBinaryOp built from vel + angVel.cross(...))
used to compute the relative velocity for the lubrication correction. Forcing
eager evaluation into concrete Vector3r variables instead of auto avoids the
bogus warning without changing behaviour, fixing the -Werror build failure.

===================================================================		 Anton Gladky <gladk@debian.org> no 2026-07-24
fix-gcc153-maybe-uninitialized-gate.patch Raise the g++ version gate for the -Wmaybe-uninitialized workaround CMakeLists.txt already disables -Werror=maybe-uninitialized below g++ 15.2.1,
on the assumption that 15.2.1 and later report it correctly. That assumption
does not hold: g++ 15.3.0 still misfires -Wmaybe-uninitialized on
boost::python::extract<T>() once inlined together with Eigen assignment
operators (confirmed empirically while fixing #1141213-style FTBFS reports;
dozens of call sites across the tree can trigger it depending on how a given
translation unit gets inlined, well beyond what's practical to patch one
call site at a time). We have no evidence any 15.x point release actually
fixes it, so rather than guess at another untested point release, gate on
the next major version (16.0.0) and revisit once that is confirmed clean.

===================================================================		 Anton Gladky <gladk@debian.org> no 2026-07-25
eigen5-longdouble-random.patch fix FTBFS with Eigen 5 on the long double flavour Eigen >= 3.4.90 picks random<Scalar>() by std::is_floating_point, which yade
specialises to true for ThinRealWrapper<long double>. That implementation needs an
unsigned integer of sizeof(long double), which Eigen has none of for 16 bytes, so it
fails with "variable or field 'randomBits' declared void". Use the custom scalar
implementation instead. Affects every architecture except amd64 and ppc64el.		 Anton Gladky <gladk@debian.org> no 2026-07-27
vtk96-check-idtype-attribute.patch make checkPotentialVTKRecorders.py tolerate the IdType attribute of VTK 9.6 VTK 9.6 writes an extra IdType="1" attribute for vtkIdTypeArray arrays. The check
splits the lines on '"' positionally, so the additional attribute shifts all the
fields and the check fails. Drop it before comparing, it says nothing about the data.		 Anton Gladky <gladk@debian.org> no 2026-07-27
fix-manpage-reproducibility-cpu-count.patch Make yade-batch.1 manpage generation reproducible help2man runs the installed yade-batch --help to build its manpage.
yade-batch.in detects the number of CPU cores from /proc/cpuinfo at
run time and embeds it verbatim into the --help text for the -j/--jobs
option ("further limited by OMP_NUM_THREADS if set by the environment:
N"). Since this text is captured at build time, the generated manpage
differs depending on how many CPUs the build machine has, making the
build non-reproducible (observed as 6 vs 4 in diffoscope between two
rebuilds of the same source on reproduce.debian.net).
.
getNumCores() returns min(int(OMP_NUM_THREADS), nCpu) when
OMP_NUM_THREADS is set in the environment, so forcing
OMP_NUM_THREADS=1 for the help2man invocation of yade-batch makes the
embedded number always evaluate to 1, independent of the build host's
actual core count. This only affects the manpage generation step, not
yade-batch's runtime behaviour.

===================================================================		 Anton Gladky <gladk@debian.org> not-needed 2026-07-30

All known versions for source package 'yade'

Links