%global commit      6565932c434827a2e2b103bb4f80e46f449cba40
%global shortcommit %(c=%{commit}; echo ${c:0:7})
%global snapdate    20251104

# CLI shape-analysis tools; strip the debug subpackage to keep it lean.
%global debug_package %{nil}

Name:           SPHARM-PDM
Version:        2.0.0
Release:        0.7.%{snapdate}git%{shortcommit}%{?dist}
Summary:        Spherical harmonic point distribution model shape analysis tools

License:        Apache-2.0
URL:            https://github.com/NIRALUser/SPHARM-PDM
Source0:        %{url}/archive/%{commit}/%{name}-%{commit}.tar.gz

BuildRequires:  gcc-c++
BuildRequires:  cmake
BuildRequires:  ninja-build
BuildRequires:  chrpath
BuildRequires:  git-core
# The three superbuild dependencies — all provided as system packages, so we
# disable the (offline-fatal) superbuild and build the inner project directly.
BuildRequires:  InsightToolkit5-devel
BuildRequires:  vtk-devel
# Defines %%vtk_abi on Fedora's vtk (via vtk-abi-compat); our vtk provides it itself.
BuildRequires:  vtk(abi)
BuildRequires:  SlicerExecutionModel-devel
# VTK cmake config runs find_package for all its module deps at configure time.
BuildRequires:  freetype-devel
BuildRequires:  pugixml-devel
BuildRequires:  fmt-devel
BuildRequires:  utf8cpp-devel
BuildRequires:  PEGTL-devel
BuildRequires:  jsoncpp-devel
BuildRequires:  nlohmann-json-devel
BuildRequires:  hdf5-devel
BuildRequires:  double-conversion-devel
BuildRequires:  expat-devel
BuildRequires:  glew-devel
BuildRequires:  lz4-devel
BuildRequires:  libxml2-devel
BuildRequires:  libtiff-devel
BuildRequires:  libjpeg-turbo-devel
BuildRequires:  libpng-devel

# Pin the exact VTK ABI this package was built against. VTK freezes its soname at
# libvtk*.so.1 across all of 9.x, so rpm cannot tell 9.5 from 9.6 from 9.7: a consumer
# built against one installs cleanly against another and fails at LOAD with undefined
# symbols. A >= floor does not help -- 9.7.0 satisfies >= 9.6.2 -- so this is an exact
# match. %%vtk_abi comes from vtk-devel's macros.vtk, so it re-pins on every rebuild.
#
# The missing-macro check lives in %%build, deliberately, NOT at parse time. A
# parse-time %%error cannot distinguish "COPR is generating the SRPM" from "this is a
# real buildroot": rpkg srpm parses the spec with no buildroot and therefore no
# vtk-devel, so a parse-time guard killed every SCM build of every guarded package
# before any chroot existed (python-simpleitk 10980296 failed in 0 minutes with
# "can't parse specfile"). Expanding to UNSET keeps the SRPM parse alive, and the
# %%build test still fails loudly in a buildroot lacking vtk-devel, so no RPM carrying
# the placeholder can ever be produced.
Requires:       vtk(abi) = %{?vtk_abi}%{!?vtk_abi:UNSET}

# itk(abi): ITK_LEGACY_REMOVE changes class layouts at unchanged sonames, so the itk_ver floor alone let a
# consumer built against the removed state install over the present one (python3-simpleitk -11 on ITK -23
# returned all-zero images). Equality on the legacy-state token; see InsightToolkit5.spec.
Requires:       itk(abi) = %{?itk_abi}%{!?itk_abi:UNSET}

%description
SPHARM-PDM is a set of command-line tools for spherical-harmonic-based shape
analysis. It computes correspondent point-based shape models (PDMs) from
binary segmentations by parameterising the surface with spherical harmonics
and sampling a corresponding icosahedral mesh, enabling group statistical
shape analysis of brain structures and other objects. Part of the NIRAL /
SlicerSALT shape-analysis ecosystem.

This build uses the system ITK, VTK and SlicerExecutionModel (no superbuild).

%prep
%autosetup -n %{name}-%{commit}

# ITK_LEGACY_REMOVE. Lifted VERBATIM from 3dslicer-spharm-pdm.spec, which is
# the same upstream project (NIRALUser/SPHARM-PDM) at a different commit --
# 6565932 here, 522de7b there -- and which took three releases discovering
# these three constructs one build at a time.
#
# This port did NOT repeat that. ci/prep-tdd.sh ran the assertions against this
# commit's unpatched source first (RED: 7 + 12 + 1 hits), then against the
# patched source (GREEN: zero). The counts match the sibling's exactly, so the
# three blocks transfer unchanged.
#
# Worth recording why the harness earned its keep: my own inventory of this
# tarball reported only TWO constructs, because the inline grep pattern left
# `()` unescaped and silently matched nothing for ComputeBoundingBox. The RED
# step found the third before a builder did.
# ITK_LEGACY_REMOVE. InsightToolkit5 5.4.7-19 ships with ITK_LEGACY_REMOVE
# defined, which deletes the class-scope aliases for the pre-5.4 enum names, so
# MeshType::CellsAllocatedDynamicallyCellByCell stopped existing and 7 call
# sites here stopped compiling. Nothing in this package changed; the buildroot
# did. That is not reversible from our side either: ITK declares
#
#   cmake_dependent_option(ITK_LEGACY_REMOVE "..." OFF "NOT ITK_WRAPPING" ON)
#
# so enabling ITK_WRAP_PYTHON FORCES it on, and -DITK_LEGACY_SILENT=ON is
# forced back off by the same mechanism. python3-itk and the legacy ITK API
# cannot both be had from one build.
#
# The modern spelling is what upstream ITK wants anyway and compiles against
# both settings, since the enum class itself is unconditional in
# itkCommonEnums.h. Worth sending to NIRAL rather than carrying forever.
python3 - <<'PYEOF'
import pathlib, re
old = re.compile(r"\b(?:Output)?MeshType::CellsAllocatedDynamicallyCellByCell\b")
new = "itk::MeshEnums::MeshClassCellsAllocationMethod::CellsAllocatedDynamicallyCellByCell"
total = 0
touched = []
for f in sorted(pathlib.Path(".").rglob("*")):
    if f.suffix not in (".cxx", ".txx", ".h", ".hxx") or not f.is_file():
        continue
    src = f.read_text(encoding="utf-8", errors="surrogateescape")
    n = len(old.findall(src))
    if n:
        f.write_text(old.sub(new, src), encoding="utf-8", errors="surrogateescape")
        total += n
        touched.append("%s (%d)" % (f, n))
assert total == 7, "expected 7 legacy enum uses, replaced %d: %s" % (total, touched)
# Zero-residual: a substitution that ran is not the same as one that finished.
left = sum(len(old.findall(f.read_text(encoding="utf-8", errors="surrogateescape")))
           for f in pathlib.Path(".").rglob("*")
           if f.suffix in (".cxx", ".txx", ".h", ".hxx") and f.is_file())
assert left == 0, "%d legacy enum uses remain" % left
print("ITK legacy enum: replaced %d uses in %d files" % (total, len(touched)))
for t in touched:
    print("   ", t)
PYEOF

# Second ITK_LEGACY_REMOVE casualty in this package (the first is the mesh
# allocation enum above). The cell-geometry enumerators were class-scope aliases
# too: CellType::LINE_CELL and friends. Found only after the first fix let the
# build get far enough to reach them -- legacy removal does not fail once, it
# fails one layer at a time, so a clean build of the first fix proved nothing
# about the second.
python3 - <<'PYEOF'
import pathlib, re
pat = re.compile(r"\bCellType::((?:VERTEX|LINE|POLYGON|TRIANGLE)_CELL)\b")
exts = (".cxx", ".txx", ".h", ".hxx")
total, touched = 0, []
for f in sorted(pathlib.Path(".").rglob("*")):
    if f.suffix not in exts or not f.is_file():
        continue
    src = f.read_text(encoding="utf-8", errors="surrogateescape")
    n = len(pat.findall(src))
    if n:
        f.write_text(pat.sub(r"itk::CommonEnums::CellGeometry::\1", src),
                     encoding="utf-8", errors="surrogateescape")
        total += n
        touched.append("%s (%d)" % (f, n))
assert total == 12, "expected 12 legacy cell-geometry uses, replaced %d: %s" % (total, touched)
left = sum(len(pat.findall(f.read_text(encoding="utf-8", errors="surrogateescape")))
           for f in pathlib.Path(".").rglob("*") if f.suffix in exts and f.is_file())
assert left == 0, "%d legacy cell-geometry uses remain" % left
print("ITK legacy cell geometry: replaced %d uses in %d files" % (total, len(touched)))
PYEOF

# THIRD ITK_LEGACY_REMOVE casualty. SpatialObject::ComputeBoundingBox() was a
# legacy shim and is gone.
#
# gcc suggests ComputeMyBoundingBox and that suggestion is WRONG, twice over.
# itkSpatialObject.h:669-677 shows what the removed method actually was:
#
#   #if !defined(ITK_LEGACY_REMOVE)
#     itkLegacyMacro(void ComputeBoundingBox())
#     {
#       this->Update(); /* Update() should be used instead of outdated ComputeBoundingBox() */
#     }
#
# so the behaviour-preserving replacement is Update(), full stop. And
# ComputeMyBoundingBox() is declared at :697, under the `protected:` at :688 --
# not callable from this call site at all. Taking the compiler's hint would
# either fail to compile or, from inside a derived class, silently compute only
# the object's own box and skip the transform/family update Update() performs.
#
# One call site. The line after it is `this->Update()` on a DIFFERENT object
# (this, vs m_CoefsMeshSpatialObject), so it is not redundant and must stay.
python3 - <<'PYEOF'
import pathlib, re
pat = re.compile(r"->ComputeBoundingBox\(\)")
exts = (".cxx", ".txx", ".h", ".hxx")
total, touched = 0, []
for f in sorted(pathlib.Path(".").rglob("*")):
    if f.suffix not in exts or not f.is_file():
        continue
    src = f.read_text(encoding="utf-8", errors="surrogateescape")
    n = len(pat.findall(src))
    if n:
        f.write_text(pat.sub("->Update()", src), encoding="utf-8", errors="surrogateescape")
        total += n
        touched.append("%s (%d)" % (f, n))
assert total == 1, "expected 1 legacy ComputeBoundingBox use, replaced %d: %s" % (total, touched)
rest = "".join(f.read_text(encoding="utf-8", errors="surrogateescape")
               for f in pathlib.Path(".").rglob("*")
               if f.suffix in exts and f.is_file())
assert "ComputeBoundingBox()" not in rest, "legacy ComputeBoundingBox() remains"
# Assert no CALL to ComputeMyBoundingBox was introduced. The declaration and
# definition are legitimate and must survive: this package OVERRIDES the
# protected virtual (SphericalHarmonicSpatialObject.{h:107,cxx:109}). That
# override is also why gcc's suggestion was dangerous rather than merely wrong
# -- ComputeMyBoundingBox() IS accessible from inside this class, so the hint
# would have compiled and silently skipped the transform update Update() does.
import re as _re
calls = _re.findall(r"[.>]ComputeMyBoundingBox\s*\(", rest)
assert not calls, "a CALL to ComputeMyBoundingBox was introduced: %r" % calls
print("ITK legacy ComputeBoundingBox -> Update: %d use in %s" % (total, touched))
PYEOF


# CMake/CMakeLists.txt does nothing but file(DOWNLOAD) a Windows-only
# FindVcvars.cmake from GitHub at configure time (fatal offline). FindVcvars is
# never used on Linux — neutralise the whole sub-CMakeLists.
echo '# FindVcvars download removed: Windows-only, unused on Linux, offline build' \
    > CMake/CMakeLists.txt

# GCC 16 rejects this legacy code's unqualified std names (e.g. MeshMath.cxx
# uses bare `cout`/`endl`, relying on a `using namespace std` that older
# standard headers pulled in transitively). Force-include a small compat header
# that brings the std names into scope for every translation unit.
cat > std-compat.h <<'EOF'
#pragma once
#include <iostream>
#include <fstream>
#include <sstream>
#include <cstdlib>
#include <cmath>
#include <cstring>
#include <string>
using namespace std;
EOF

# Fedora's vtk (9.7, rawhide) builds the Java wrapping and exports VTK::Java with
# the JDK's /usr/lib/jvm/java/lib/libjava.so in its link interface. SPHARM-PDM links
# VTK modules that pull VTK::Java in, so that file lands on the link line: without
# a JDK ninja has no rule for it, and with one ld cannot resolve its libjvm.so.
# SPHARM-PDM never calls into Java: right after find_package(VTK) in
# SPHARM-PDM.cmake, drop the JDK library from the imported target, and once every
# directory is configured take VTK::Java off the link lists of every target
# SPHARM-PDM defines (same fix as 3dslicer.spec).
python3 - SPHARM-PDM.cmake <<'PYEOF'
import re, sys
f = sys.argv[1]; s = open(f).read()
inj = r'''
if(TARGET VTK::Java)
  get_target_property(_pkgvtk_java VTK::Java INTERFACE_LINK_LIBRARIES)
  list(FILTER _pkgvtk_java EXCLUDE REGEX "/libjava[.]so$")
  set_property(TARGET VTK::Java PROPERTY INTERFACE_LINK_LIBRARIES "${_pkgvtk_java}")
  function(_pkgvtk_drop_java _dir)
    get_property(_tgts DIRECTORY "${_dir}" PROPERTY BUILDSYSTEM_TARGETS)
    foreach(_t IN LISTS _tgts)
      foreach(_p LINK_LIBRARIES INTERFACE_LINK_LIBRARIES)
        get_property(_l TARGET ${_t} PROPERTY ${_p})
        if(_l MATCHES "VTK::Java")
          list(FILTER _l EXCLUDE REGEX "VTK::Java([^A-Za-z0-9_]|$)")
          set_property(TARGET ${_t} PROPERTY ${_p} "${_l}")
        endif()
      endforeach()
    endforeach()
    get_property(_subs DIRECTORY "${_dir}" PROPERTY SUBDIRECTORIES)
    foreach(_s IN LISTS _subs)
      _pkgvtk_drop_java("${_s}")
    endforeach()
  endfunction()
  cmake_language(DEFER CALL _pkgvtk_drop_java "${CMAKE_CURRENT_SOURCE_DIR}")
endif()
'''
m = re.search(r'^[ \t]*find_package[ \t]*\([ \t]*VTK\b[^)]*\)[^\n]*\n', s, re.M | re.I)
if not m: sys.exit("find_package(VTK) not found in " + f)
open(f, 'w').write(s[:m.end()] + inj + s[m.end():])
PYEOF
grep -q '_pkgvtk_drop_java' SPHARM-PDM.cmake

%build
# Pin the ITK layout state this build sees (see the Requires above).
test -n "%{?itk_abi}" || { echo "ERROR: itk_abi undefined - needs InsightToolkit5-devel >= 5.4.7-24 (macros.itk)" >&2; exit 1; }
# vtk_abi comes from vtk-devel's macros.vtk. Absent means BuildRequires: vtk-devel is
# missing, and the Requires above would ship as the vacuous "vtk(abi) = UNSET".
test -n "%{?vtk_abi}" || { echo "ERROR: vtk_abi undefined - add 'BuildRequires: vtk-devel'" >&2; exit 1; }
export CXXFLAGS="%{optflags} -include ${PWD}/std-compat.h -fpermissive"
%cmake -GNinja \
    -DCMAKE_BUILD_TYPE=Release \
    -DSPHARM-PDM_SUPERBUILD:BOOL=OFF \
    -DUSE_SYSTEM_ITK:BOOL=ON \
    -DUSE_SYSTEM_VTK:BOOL=ON \
    -DUSE_SYSTEM_SlicerExecutionModel:BOOL=ON \
    -DBUILD_TESTING:BOOL=OFF \
    -DCMAKE_SKIP_INSTALL_RPATH:BOOL=ON
# %%prep strips the JDK's libjava.so from VTK::Java and VTK::Java from every target;
# if either still reaches a link line, stop here instead of hours into the build.
if grep -qE '/libjava[.]so|libvtkJava[.]so' %{_vpath_builddir}/build.ninja; then
    echo 'ERROR: libjava.so or libvtkJava.so is on a link line; the VTK::Java fixes in %%prep did not apply' >&2
    exit 1
fi
%cmake_build

%install
%cmake_install

# This is a CLI-tools package: drop the static dev libraries and headers the
# build also installs (the tools are statically linked and self-contained).
rm -rf %{buildroot}%{_includedir}/Shape
find %{buildroot} -name '*.a' -delete

# Strip any residual build-tree RPATHs the SEM macros may leave.
find %{buildroot} -type f -perm -111 \
    -exec chrpath --delete {} \; 2>/dev/null || :

%files
%license License.txt
%{_bindir}/*

%changelog
* Fri Oct 30 2026 Morgan Hough <morgan.hough@gmail.com> - 2.0.0-0.7.20251104git6565932
- Fedora's vtk 9.7 (rawhide) exports VTK::Java with the JDK's libjava.so in its link
  interface; the rawhide build failed at the first link ("libjava.so, needed by
  bin/STL2Meta, missing and no known rule"). %%prep now strips libjava.so from
  VTK::Java right after find_package(VTK) in SPHARM-PDM.cmake and, via
  cmake_language(DEFER), drops VTK::Java from every target's link lists (the
  3dslicer.spec fix). %%build stops after configure if libjava.so or libvtkJava.so
  is still in build.ninja.

* Tue Oct 06 2026 Morgan Hough <morgan.hough@gmail.com> - 2.0.0-0.6.20251104git6565932
- Pin the ITK layout state: Requires: itk(abi) = %%{itk_abi}, from InsightToolkit5-devel's macros.itk,
  plus a %%build guard. ITK -23 flipped ITK_LEGACY_REMOVE back off, which changes class layouts at
  unchanged sonames; this package was built against -10..-21 (legacy removed) and the itk_ver floor let it
  install over -23, giving silent wrong results or crashes. Rebuild against ITK >= 5.4.7-24.

* Mon Oct 05 2026 Morgan Hough <morgan.hough@gmail.com> - 2.0.0-0.5.20251104git6565932
- Port 3dslicer-spharm-pdm.spec's three ITK_LEGACY_REMOVE blocks. Same upstream
  project at a different commit (6565932 here, 522de7b there), and the
  construct counts match exactly: 7 mesh-allocation enums, 12 cell-geometry
  enums, 1 SpatialObject::ComputeBoundingBox call.
- Surfaced by medInria failing on fedora-rawhide-x86_64, where nothing provides
  libitkvnl-5.4.so.1: InsightToolkit5 5.4.7-21 moved to the system vxl and
  ships no bundled vnl. Eight packages on that chroot still link it and every
  one carries an .fc45 tag -- builds never redone for fc46. Only three are
  genuinely stale rather than superseded leftovers, and this is one.
- Verified with ci/prep-tdd.sh rather than by building: RED on unpatched source
  (7 + 12 + 1), GREEN after. That caught an error in my own inventory, which
  reported only two constructs because an inline grep left `()` unescaped and
  matched nothing for ComputeBoundingBox. The sibling spent three builds
  finding its three layers serially; this found all three in about thirty
  seconds.
* Sat Sep 19 2026 Morgan Hough <morgan.hough@gmail.com> - 2.0.0-0.4.git
- Rebuild against InsightToolkit5 5.4.7-8, which unbundles Eigen3, VXL and
  double-conversion. No source or spec change; the Release bump is what makes
  the rebuild reach anyone. COPR adds a sibling at an already-published NVR
  rather than replacing it, and dnf then serves the older copy -- so a rebuild
  at the same NVR would leave this package linked against the bundled ITK while
  appearing to succeed.
- Validated against -8 in mhough/neurofedora-itk5-test before this bump.
* Sun Aug 23 2026 Morgan Hough <morgan.hough@gmail.com> - 2.0.0-0.2.20251104git6565932
- Rebuild against VTK 9.7.0. VTK here installs unversioned sonames
  (libvtk*.so.1), so 9.7 replaces 9.6 in place and every consumer must be
  rebuilt or it fails at runtime with undefined symbols.

* Mon Jun 15 2026 Morgan Hough <morgan.hough@gmail.com> - 2.0.0-0.1.20251104git6565932
- Initial package: SPHARM-PDM spherical-harmonic shape-analysis CLI tools
  (master snapshot 6565932). Built against system ITK/VTK/SlicerExecutionModel
  with the superbuild disabled (Fedora no-superbuild policy). Complements the
  FSL FIRST mesh shape-analysis workflow.