Rewrite README: correct install instructions, document what ships

[gdevenyi]

The README's install section had drifted from the build system. Several documented options no longer exist, so following the README as written produces a configure line with dead flags and a wrong picture of what is bundled.

Everything below was checked against CMakeLists.txt, cmake-modules/, .github/workflows/, and an install tree.

Correctness fixes

Options that do not exist

• -DMT_BUILD_IM — no such option anywhere in the project.
• -DUSE_SYSTEM_PCRE — PCRE was removed in 886ec1a (Drop vendored PCRE; oobicpl now uses C++ std::regex). There is no option, no BuildPCRE.cmake, and no binary links it. PCRE was also still listed under "built internally".
• -DUSE_SYSTEM_GLUT=OFF — misleading. MT_USE_GLFW defaults to ON, so GLUT is not used at all unless GLFW is turned off.

Wrong versions

• CMake minimum is 3.10, not 3.1.
• ITK is 4.14 (pinned commit cae3eb9). The README said 4.11 in one list and 4.13 in another.
• The prose said the example installs to /opt/minc/1.9.17; the command said 1.9.18.

Environment setup was backwards

The old text said to source minc-toolkit-config.sh only if you used a non-default prefix. It is required in every shell, at any prefix, and the README never gave its path.

Markup — the </pre></code> closing tag was malformed and prose was mixed inside the code blocks.

New material

• Download section — the .deb/.rpm/.pkg artifacts, the distro matrix, and minimal vs full, taken from release.yml. The README previously had no mention that binary packages exist.
• Homebrew — the vfonov/minc tap as a macOS alternative.
• Cloning — git submodule update --init --recursive for an existing non-recursive clone, plus a branch/tag table. Submodules being mandatory was never stated.
• Option tables with real defaults, including the previously undocumented MT_USE_BLAS, MT_BUILD_OPENBLAS, BLAS_PREFERENCE, MT_USE_GLFW, MT_BUILD_ABC, MT_BUILD_DCM2MNC_TESTS, and the JPEG/OpenJPEG/libarchive/PNG/expat system options.
• Dependencies matched to CI. The old lists omitted GSL, FFTW, NetCDF, HDF5, OpenJPEG, libarchive, expat, OpenBLAS, gfortran, ImageMagick and GLFW. CentOS 7 is EOL, so it is replaced by Fedora/RHEL and macOS.
• What is installed — the install tree layout, programs grouped by job, the pipeline scripts, and the data in share/ and etc/.
• Data files that are not included — the two things that make a tool fail out of the box: the stereotaxic models mritotal looks for in share/mni-models, and the BEaST prior library in share/beast-library-1.0.

Verification

Program names come from an install tree or from the source. Option names and defaults come from CMakeLists.txt and cmake-modules/. Artifact names and the distro matrix come from release.yml. Dependency lists come from ci.yml.

Documentation only. No build files touched.

🤖 Generated with Claude Code

https://claude.ai/code/session_019katBUnczdvUS7WQ4RpVoa

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[gdevenyi]

Added a commit here because #230 invalidates one line of this rewrite.

The note under the USE_SYSTEM_* table currently reads:

USE_SYSTEM_NIFTI has no option() entry, so it does not show up in ccmake. Passing it on the command line still works.

#230 makes it a real option(), so that stops being true. Replaced it with the thing actually worth documenting — the trade-off:

USE_SYSTEM_NIFTI carries one caveat. The bundled NIfTI is built from a pinned nifti_clib with every exported symbol renamed to minc_*, so that it cannot collide with the unmangled niftiio that ITK bundles (ITK 4.x has no ITK_USE_SYSTEM_NIFTI switch). A system NIfTI is unmangled. That is fine against the shared libniftiio.so distributions package — which is what the release .deb and .rpm link — but do not combine USE_SYSTEM_NIFTI with a static system NIfTI and MT_BUILD_ITK_TOOLS, or the two copies will collide on nifti_image_read and friends at link time.

The table row itself is unchanged. If #230 lands first this is already accurate; if this lands first the note describes behaviour that arrives one PR later, so #230 is the one to merge first of the two.

[gdevenyi]

Rebased onto develop-1.9.18 at 342c896. One conflict, in the section this PR replaces: #222 removed arguments from the old bundled-packages list while this rewrite replaces that list wholesale, so the rewrite wins.

That merge does leave one stale row in the new package table, fixed as part of the rebase — the rewrite still listed arguments, which #222 replaced with cxxopts:

| [cxxopts](https://github.com/jarro2783/cxxopts) | Header-only command-line argument parsing, used by oobicpl. |

Now MERGEABLE again.