tests: remove fifteen #define private public — by deleting dead hacks and by passing INPUT values explicitly

[Critsium-xy]

What this does

Removes 15 of the 69 #define private public / #define protected public
macros in source/ (69 → 54), in two commits. No friend declaration is
added anywhere — every macro is removed either because it was never needed, or
by making the dependency explicit, which is why the diff reduces global-state
coupling instead of relocating it.

PARAM/GlobalV/GlobalC ledger for the whole PR: added 5, removed 31,
net_delta = −26.

---

Commit 1 — five macros no test actually needed

Four were vestigial: the tests inside them access no private or protected member
of any class in the headers the macro covers.

File	Why the macro could go
source_cell/test/klist_test_para.cpp	The only mentions of K_Vectors::spin_mult and mpi_k() are in comments. koffset at lines 209/333 is a local const double koffset[3], not the private member of the same name.
source_cell/test/qlist_test.cpp	Every member it reads — nkstot, nkstot_nospin, wk, kvec_c, kvec_d, kc_done, kd_done, is_mp — is public in ModuleCell::ReciprocalGrid. QList's own privates (nirr_, irrep_modes_, little_group_, interpolate_q_between, get_irreps) are never touched.
source_cell/test/sepcell_test.cpp	Already goes through the public getters get_ntype(), get_omega(), get_tpiba2(), get_sep_enable(), get_seps(). sep_enable appears only inside a comment.
source_lcao/test/test_output_hcontainer_consistency.cpp	Uses none of the private members of HContainer, Output_HContainer or Read_HContainer.
source_lcao/test/test_init_dm_from_file.cpp	Read DensityMatrix::_DMR directly in 7 places while calling the public get_DMR_vector() in the same tests. get_DMR_vector() returns exactly _DMR (density_matrix.h:191), so those 7 reads now go through it.

The three source_cell macros also covered pseudo.h, atom_spec.h,
magnetism.h and atom_pseudo.h, which have no private sections at all —
including them under the macro never did anything.

Commit 2 — ten macros that existed only to write the private half of PARAM

Group A: the dependency was already injected; the test routed through the global for no reason

No production change at all in this group.

• propagator_test{1,2,3} (6 macros) — Propagator's constructor already
  takes const double& dt. Each test wrote PARAM.input.mdp.md_dt and then read
  PARAM.mdp.md_dt straight back to pass it in. Now a local md_dt. The
  now-unused parameter.h include is dropped; propagator.h already provides
  ModuleBase::AU_to_FS via source_base/constants.h, and
  source_base/global_variable.h is included directly for the
  GlobalV::ofs_running these tests genuinely use.
• single_r_io_test (1) — PARAM.sys.nlocal = 99 was dead.
  single_r_io.cpp has no PARAM reference at all and takes nlocal from
  pv.get_global_row_size(), which this test stubs to return 5.
• read_wfc_nao_test (1) — read_wfc_nao() already takes the directory as
  its first argument; the test wrote PARAM.sys.global_readin_dir and passed
  it straight back in. Now a local readin_dir.

Group B: the production code really did read PARAM, so the value became a parameter

• write_dmk() now takes const std::string& dmk_dir, mirroring its sibling
  read_dmk() in the same header, which already does. write_dmk.cpp is now
  PARAM-free and drops the include. Its one caller
  (ctrl_scf_lcao.cpp:199) already holds a global_out_dir local, so the call
  site adds no PARAM reference.
• K_Vectors::set_spin_mult() added next to the existing set_nks() /
  set_nkstot() / set_nkstot_nospin(). The public getter get_spin_mult()
  already existed, so this completes an incomplete setter group and lets
  write_dmk_test stop assigning kv.spin_mult directly — the second reason
  that file needed the macro.
• write_eig_iter() and write_eig_file() take nbands and nspin, and
  write_eig_file() takes const std::string& out_dir. Both callers are in
  esolver_ks.cpp, which already holds an injected inp_, so nbands/nspin
  cost no PARAM reference.
• write_eig_occ_test's PARAM.sys.nbands_l write was dead (no source or
  library in that target reads it — the readers are in source_estate,
  source_psi and source_pw), and its PARAM.input.bndpar read is now a
  local mirroring the Input_para default of 1, so the test no longer depends
  on an INPUT default.

No default arguments were added (governance blocks that, and print_force
already showed why: adding one is not an option). No assertion or expected
value changed.

Governance warnings, with rationale

• Global dependency budget — 5 added vs 31 removed, net_delta = −26.
  • esolver_ks.cpp:328 (+1) is the only genuinely new reference:
    PARAM.globalv.global_out_dir at the write_eig_file call site. This is the
    composition root deliberately absorbing a global read so the leaf writer stays
    clean — the same direction as source_relax: pass INPUT values explicitly instead of reading PARAM, removing the test access hacks it caused #7921.
  • write_eig_occ_test.cpp:66,67 (+4) are pre-existing GlobalV::NPROC /
    GlobalV::MY_RANK arguments to Parallel_Global::init_pools; reflowing that
    call to pass a local bndpar makes the checker count those lines as added.
• Header dependency — write_eig_occ.h gains <string>, required by the new
  const std::string& out_dir parameter. No forward declaration is possible for
  std::string here.
• Documentation — no INPUT parameter or user-facing behaviour changed, so no
  docs/ update is required and docs/parameters.yaml needs no regeneration.

Not in this PR

• source_io/module_json/test/para_json_test.cpp — reads the private static
  Json::AbacusJson::doc 89 times (77 as doc[...] subscripts). Its only
  public path is write_to_json(filename). Also note this target builds only
  under ENABLE_RAPIDJSON.
• source_estate/module_dm/test/test_dm_r_init.cpp — asserts on
  DensityMatrix::_DMR_save[0].size() in 2 places; needs a production accessor.
• source_io/test/outputlog_test.cpp — out_level and nspin are cheap
  (1 call site each), but its third key test_force lives in print_force,
  which has ~45 call sites across force_pw.cpp, sto_forces.cpp,
  force_stress_assemble.cpp, md_func.cpp and esolver_{dp,lj,nep}.cpp.
  Injecting it would add ~45 PARAM references and block on the budget, all
  for one std::cout << table; debug echo. It needs its own PR paired with a
  real reduction elsewhere.

Continues #7921 / #7949 / #7952 / #7953 / #7963 / #7964 / #7965 / #7966 / #7967
and keeps their scope rule: a file is either taken fully off the macro or left
alone.

Verification

Remote Linux box, cmake -B build -G Ninja -DBUILD_TESTING=ON -DENABLE_LCAO=ON -DENABLE_MPI=ON -DENABLE_OPENMP=ON:

• configure exit 0; 3354/3354 targets, build exit 0, zero errors,
  ninja: no work to do on a follow-up full build
• every changed file confirmed compiled, and all affected targets link:
  MODULE_CELL_klist_test_para1, MODULE_CELL_qlist_test,
  MODULE_CELL_SEP_CELL_TEST, MODULE_LCAO_init_dm_from_file_test,
  MODULE_LCAO_output_hcontainer_consistency_test,
  MODULE_LCAO_tddft_propagator_test, MODULE_IO_single_R_test,
  MODULE_IO_read_wfc_nao_test, MODULE_IO_write_dmk,
  MODULE_IO_write_eig_occ_test, plus abacus_basic_para
• agent_governance_check.py --base upstream/develop --head HEAD: zero
  blockers, no access-hack ratchet finding (15 removed, 0 added)

🤖 Generated with Claude Code

[mohanchen]

LGTM