@@ -10182,46 +10182,7 @@ packages:
   repo: https://github.com/fplll/fpylll
   status: ci-running
   pr: https://github.com/riseproject-dev/python-wheels/pull/3032
-  notes: 'PR #3032 open, CI running. Cython bindings over the fplll C++ lattice-reduction
-    library (LLL/BKZ). fpylll''s own setup.py expects a system-installed fplll
-    (libraries=["gmp","mpfr","fplll"], no vendoring) -- its tests.yml/sdist.yml
-    build fplll from a plain git clone via autogen.sh+configure+make install
-    (--disable-static --with-max-enum-dim=64 --with-max-parallel-enum-dim=20);
-    its release.yml (what actually produces upstream''s PyPI wheels) instead
-    checks out passagemath/passagemath and uses its Sage-distribution SPKG
-    tooling to build gmp/mpfr/mpc/fplll, which has no riscv64 equivalent here,
-    so tests.yml''s from-source shape is what this port mirrors. GMP/MPFR
-    confirmed in Rocky 10 riscv64''s own repos (gmp/gmp-devel/gmp-c++ in
-    BaseOS/AppStream, mpfr/mpfr-devel in BaseOS/AppStream via primary.xml --
-    same libmpc-devel finding already recorded for the passagemath-objects/
-    epicscorelibs entries), confirmed via filelists.xml to land in /usr/lib64
-    (gmp-devel: libgmp.so, mpfr-devel: libmpfr.so, both with
-    /usr/share/licenses/<pkg>/ texts). fplll itself has no Rocky package on
-    any arch (CRB/AppStream/BaseOS all checked) and is built once per
-    container from its own 5.5.0 GitHub release tarball (sha256-pinned; that
-    tarball ships a pregenerated ./configure, unlike GitHub''s raw source
-    snapshot, so no autoreconf/autoconf/automake/libtool needed) with
-    --prefix=/usr --libdir=/usr/lib64 so it lands next to gmp/mpfr with no
-    extra -L/-I/LD_LIBRARY_PATH plumbing for setup.py''s hardcoded
-    sys.prefix-relative include/library dirs. libqd (quad-double) skipped:
-    Rocky ships no qd package on any arch, and HAVE_QD=no is one of the two
-    configurations upstream''s own tests.yml matrix already exercises, not a
-    reduced-feature workaround. Checked fplll''s full C++ core (configure.ac
-    and src/) for ISA-gated/SIMD code (AVX/SSE/NEON/immintrin/__builtin_cpu/
-    march=native/etc.) -- zero hits; the library has no vectorized
-    inner-product fast path to gate. cysignals (build+runtime dep, cimported
-    in every .pyx) and numpy (optional HAVE_NUMPY extension + test dep) both
-    already have riscv64 wheels on this registry for cp312/cp313/cp314/cp314t,
-    so the matrix matches that range (cysignals has no cp311 build here
-    either, matching gmpy2/symengine). musllinux dropped, same as
-    symengine/gmpy2/chiavdf (no riscv64 musl rehearsal yet for this
-    from-source-C++-dependency shape). gpl_sources job collects gcc/gmp/mpfr
-    (vendored RPM-built shared libs + the image''s own gcc), matching
-    symengine''s convention exactly. Local build rehearsal (docker) was not
-    possible in this sandbox: the docker daemon cannot start here (no
-    systemd, ulimit/cgroup restrictions) and disk was tight (~8GB free)
-    besides -- relying on the PR''s own pull_request-triggered riscv64 CI run
-    instead.'
+  notes: 'PR #3032 open, CI running. Cython bindings over the fplll C++ lattice-reduction library (LLL/BKZ). fpylll''s own setup.py expects a system-installed fplll (libraries=["gmp","mpfr","fplll"], no vendoring) -- its tests.yml/sdist.yml build fplll from a plain git clone via autogen.sh+configure+make install (--disable-static --with-max-enum-dim=64 --with-max-parallel-enum-dim=20); its release.yml (what actually produces upstream''s PyPI wheels) instead checks out passagemath/passagemath and uses its Sage-distribution SPKG tooling to build gmp/mpfr/mpc/fplll, which has no riscv64 equivalent here, so tests.yml''s from-source shape is what this port mirrors. GMP/MPFR confirmed in Rocky 10 riscv64''s own repos (gmp/gmp-devel/gmp-c++ in BaseOS/AppStream, mpfr/mpfr-devel in BaseOS/AppStream via primary.xml -- same libmpc-devel finding already recorded for the passagemath-objects/ epicscorelibs entries), confirmed via filelists.xml to land in /usr/lib64 (gmp-devel: libgmp.so, mpfr-devel: libmpfr.so, both with /usr/share/licenses/<pkg>/ texts). fplll itself has no Rocky package on any arch (CRB/AppStream/BaseOS all checked) and is built once per container from its own 5.5.0 GitHub release tarball (sha256-pinned; that tarball ships a pregenerated ./configure, unlike GitHub''s raw source snapshot, so no autoreconf/autoconf/automake/libtool needed) with --prefix=/usr --libdir=/usr/lib64 so it lands next to gmp/mpfr with no extra -L/-I/LD_LIBRARY_PATH plumbing for setup.py''s hardcoded sys.prefix-relative include/library dirs. libqd (quad-double) skipped: Rocky ships no qd package on any arch, and HAVE_QD=no is one of the two configurations upstream''s own tests.yml matrix already exercises, not a reduced-feature workaround. Checked fplll''s full C++ core (configure.ac and src/) for ISA-gated/SIMD code (AVX/SSE/NEON/immintrin/__builtin_cpu/ march=native/etc.) -- zero hits; the library has no vectorized inner-product fast path to gate. cysignals (build+runtime dep, cimported in every .pyx) and numpy (optional HAVE_NUMPY extension + test dep) both already have riscv64 wheels on this registry for cp312/cp313/cp314/cp314t, so the matrix matches that range (cysignals has no cp311 build here either, matching gmpy2/symengine). musllinux dropped, same as symengine/gmpy2/chiavdf (no riscv64 musl rehearsal yet for this from-source-C++-dependency shape). gpl_sources job collects gcc/gmp/mpfr (vendored RPM-built shared libs + the image''s own gcc), matching symengine''s convention exactly. Local build rehearsal (docker) was not possible in this sandbox: the docker daemon cannot start here (no systemd, ulimit/cgroup restrictions) and disk was tight (~8GB free) besides -- relying on the PR''s own pull_request-triggered riscv64 CI run instead. GREEN 2026-10-15: all 13 CI checks on PR #3032 passed - setup, check_patches/check_commit_messages/check_packages_yaml, all four build legs (cp312/cp313/cp314/cp314t), Collect GPL sources, and the Publish job itself all completed with conclusion success. No open review threads. Undrafted the PR (per standing rule: mark ready for review once CI is fully green). Not merged - no separate explicit authorization for that. pypi.riseproject.dev still 404s and docs/packages/fpylll.yaml doesn''t exist on main yet - normal async pipeline lag (the publish job only just completed, the docs-update bot commit + website.yml redeploy haven''t landed yet), not a stuck/broken state - no action needed unless it''s still missing after the usual propagation delay.'
 - pkg: triton-ascend
   version: 3.2.0
   home: https://gitcode.com/Ascend/triton-ascend/