Complete warning-free step 2.3A cleanup

ALPS3_MIGRATION_PLAN.md

@@ -361,29 +361,62 @@ and Boost.Math in `legendre_convert`.
 
 ### 2.3 Code fixes
 
-* Fix B1 through B20 (§1.6), including B7b. B21 (logging convention and verbosity) can go here or into a later step; it changes the CLI regression references on purpose.
-* Moved here from 2.2 (they change the CLI references on purpose): replace
-  `boost::diagnostic_information` by `e.what()` (with B1), add a `SEED`
-  parameter for the bootstrap (B10), and delete the commented-out ublas and
-  LAPACK-bindings code.
-* `legendre_convert` (decided 2026-09-24), in this order:
-  1. Add regression cases for `legendre_convert` first (it has no tests), with
-     references from the current Boost build, like the `kk` cases in 2.2.
-  2. Replace Boost.Random by `<random>`. The `mt19937` engine gives identical
-     numbers; the normal variates differ (different algorithm), which is
-     harmless because the error estimate is seeded from the clock.
-  3. Replace `boost::math::factorial` by a product (only small arguments
-     occur).
-  4. Replace `boost::math::legendre_p` by the standard three-term recurrence.
-     `std::legendre` is not an option: libc++ (AppleClang) does not implement
-     the C++17 special math functions.
-  Keep `boost::math::sph_bessel` (also missing in libc++; our own version
-  would need careful checking at large l and argument) and `program_options`
-  (no standard equivalent). B5, B6 and B7 change reachable behavior, so each gets its own commit with a before/after test.
-* Remove all `using namespace boost::numeric;` lines, `#include <alps/config.hpp>`, dead commented-out ublas and lapack-bindings code, and the unused `alps::cast`.
-* Make `eigen_hdf5.hpp`/`eigen_lapack.hpp` functions `inline`, or move them into `.cpp` files (B8).
-* Use `Eigen::Index` for loop indices (B14). Consider `BDCSVD` in place of `JacobiSVD` (B11); that one is a numerics change and needs checking against the references.
-* Keep the `MaxEntSimulation` public getters stable, because the tests use them.
+Grouped by what each item does to the regression references (agreed
+2026-09-24). One PR per group, in this order, each with a Copilot review.
+Already done elsewhere: B2, B3 (2.1), B7 (`xi_bose_bugfix`), B20 (Pade removed).
+
+**2.3A: no change to results** (references must stay bit-identical)
+
+**Done 2026-09-24.** All Maxent-owned code and utilities build with
+`-Wall -Wextra -Wpedantic -Werror`; the fast and full regression suites remain
+unchanged. A text-input regression covers trailing blank lines.
+
+* Fix all compiler warnings (44 at the start of 2.3): B14 (`std::size_t` vs
+  `Eigen::Index` loops), B12 (member initialization order), unused variables,
+  misleading indentation, B13 (deprecated `params::get_origin_name()`; the
+  default `BASENAME` must not change). Then turn on `MAXENT_WERROR` in the `dev`
+  preset.
+* B8: make the functions in `eigen_hdf5.hpp`/`eigen_lapack.hpp` `inline`.
+* B4: make the size check in `Backcont::max_error` unconditional, avoiding an
+  out-of-bounds access in every build type.
+* B5, B6: remove the unreachable kernel branches.
+* B9: the data-file reading loop (`while (datstream)` reads past EOF); same
+  results for valid files, plus a test with trailing blank lines.
+* Remove the commented-out ublas and LAPACK-bindings code, `<alps/config.hpp>`
+  includes and the unused `alps::cast`.
+
+**2.3B: `legendre_convert`** (decided 2026-09-24), in this order:
+
+1. Add regression cases for `legendre_convert` first (it has no tests), with
+   references from the current Boost build, like the `kk` cases in 2.2.
+2. Replace Boost.Random by `<random>`. The `mt19937` engine gives identical
+   numbers; the normal variates differ (different algorithm), which is
+   harmless because the error estimate is seeded from the clock.
+3. Replace `boost::math::factorial` by a product (only small arguments occur).
+4. Replace `boost::math::legendre_p` by the standard three-term recurrence.
+   `std::legendre` is not an option: libc++ (AppleClang) does not implement the
+   C++17 special math functions.
+
+Keep `boost::math::sph_bessel` (also missing in libc++; our own version would
+need careful checking at large l and argument) and `program_options` (no
+standard equivalent).
+
+**2.3C: command-line behavior** (updates the CLI references on purpose)
+
+* B1: non-zero exit code on errors, together with `boost::diagnostic_information`
+  → `e.what()` (moved here from 2.2).
+* B16: accept `half-lorentzian` as well as `half lorentzian`, and show the real
+  `CUT` default in `--help.grids`.
+* B10: a `SEED` parameter for the bootstrap errors (moved here from 2.2).
+
+**2.3D: numerical changes** (each its own commit and `REFERENCE_CHANGES.md` entry)
+
+* B7b: time-bosonic kernel: use the ω→0 limit only at ω = 0.
+* B15: remove the `float` casts in the log grid.
+* B11 (optional): `BDCSVD` instead of `JacobiSVD`; measure against the
+  references before deciding.
+
+Keep the `MaxEntSimulation` public getters stable, because the tests use them.
 
 ### 2.4 Hygiene and CI
 
@@ -406,6 +439,17 @@ This is the last task of step 2, still on ALPSCore:
 2. **HDF5 seam:** move all archive access into `maxent_io_hdf5.{hpp,cpp}`, with functions such as `write_vector(path, Eigen::VectorXd)`, `read_vector(path)` and `write_params(...)`. Nothing else includes `<alps/hdf5...>`.
 3. Tests use `std::filesystem::temp_directory_path()` in place of `alps::temporary_filename`.
 
+### 2.6 Physics and algorithms (after 2.3–2.5)
+
+Moved out of 2.3 (2026-09-24) because they are physics or algorithm changes
+rather than cleanup; each needs its own decision:
+
+* B17: Legendre bosonic kernel (currently the fermionic formula).
+* B18: default grid when `OMEGA_MIN=0` (change the default, or warn).
+* B19: minimizer divergence (step-size control or a trust region).
+* B21: logging convention and verbosity (issue #46).
+* B22: performance of the core at large NFREQ.
+
 **Step 2 acceptance:**
 
 * Configures with CMake 3.22 through 4.x.
@@ -640,3 +684,4 @@ make -j8 && ctest     # 6/6 executables, 35 cases pass, ~9 s
 | 2026-09-23 | 2.x | Pade removed (D7), on branch `modernize/remove-pade`: `pade/` (13 files: 6 built, 5 unbuilt alternatives, header, CMake), `MAXENT_BUILD_PADE`, and the README section. It required GMP (`mpf_class`, 256-bit default precision) and did not compile (B20). |
 | 2026-09-24 | 2.2 | Step 2.2 done on `modernize/step2.2`: kk regression cases, kk spline, closed-form Legendre kernel (GSL removed), Boost utilities → std. No reference changed; 41/41 tests. ALPSCore#668 merged (GoogleTest 1.18, not installed, CMake 3.16). Found B22 (performance). |
 | 2026-09-24 | sync | Synced with ALPSCore `master` `560ae112` (#669: C++14 required and default, ALPSCore compiled with exactly `ALPS_CXX_STD`, Boost.Math dependency removed; #670: Doxygen removed). Maxent `d529a14` against it (installed as C++14, Boost 1.88): configures without `Boost_DIR`, compiles as C++17, 41/41 tests, regression unchanged (only the known Legendre rounding-level differences from 2.2). Note: build ALPSCore in a fresh build directory; an old one keeps a stale compiler identification that Maxent's configure then rejects. |
+| 2026-09-24 | 2.3A | Warning cleanup and no-output-change fixes complete: Eigen index types, initializer order, unused/deprecated code, inline HDF5/LAPACK helpers, unconditional size validation, unreachable kernel branches, robust three- and five-column text-file extraction, and dead bindings/config code removed. The dev preset enables `MAXENT_WERROR`; 44/44 Release tests pass including full regression, and the LAPACK build is warning-free. |

CMakePresets.json

@@ -20,7 +20,10 @@
       "name": "dev",
       "displayName": "Debug",
       "inherits": "base",
-      "cacheVariables": {"CMAKE_BUILD_TYPE": "Debug"}
+      "cacheVariables": {
+        "CMAKE_BUILD_TYPE": "Debug",
+        "MAXENT_WERROR": "ON"
+      }
     },
     {
       "name": "asan",

legendre_convert/legendre_convert.cpp

@@ -324,7 +324,7 @@ double GtBoot(vector_type gl_in, void *arg){
 //************************
 int main(int argc, char**argv){
 	namespace po = boost::program_options;
-	double beta,c1,c2,c3;
+	double beta,c1,c2;
 	int lmax,numConvergence,maxit,maxn;
 	bool backcontinue = false,enforceTailOff = false,enforceErrCheck=true,continueMatsubara=false;
 	std::string input_gtau_filename,gl_filename;

src/eigen_hdf5.hpp

@@ -11,7 +11,6 @@
 
 #pragma once
 #include <alps/hdf5/archive.hpp>
-#include <alps/utilities/cast.hpp>
 #include <Eigen/Core>
 #include <algorithm>
 
@@ -32,7 +31,7 @@ namespace alps {
                       if (value.size()) {
                           std::vector<std::size_t> first(get_extent(value[0]));
                           if (true)
-                              for(std::size_t i=1;i<value.size();i++) {
+                              for(Eigen::Index i=1;i<value.size();i++) {
                                   std::vector<std::size_t> size(get_extent(value[i]));
                                   if (
                                          first.size() != size.size()
@@ -46,15 +45,14 @@ namespace alps {
                   }
           };
         }
-        void save(
+        inline void save(
             archive &ar
           , std::string const & path
           , Eigen::VectorXd &value
           , std::vector<std::size_t> size = std::vector<std::size_t>()
           , std::vector<std::size_t> chunk =std::vector<std::size_t>()
           , std::vector<std::size_t> offset = std::vector<std::size_t>()
        ) {
-            using alps::cast;
             if (ar.is_group(path))
                 ar.delete_group(path);
             if (is_continuous<Eigen::VectorXd>::value && value.size() == 0)

src/eigen_lapack.hpp

@@ -21,9 +21,7 @@ extern "C" void dgesvd_( const char* jobu, const char* jobvt,
 ///performs a SVD on input matrix K
 /// returns K =  U (S) V^T where S is a vector of the diagonal values of
 /// matrix Sigma
-void lapack_svd(matrix_type &K, vector_type &S, matrix_type &Vt, matrix_type &U){
-
-   /*boost::numeric::bindings::lapack::gesvd('S', 'S', Kt, S, U_, Vt_);*/
+inline void lapack_svd(matrix_type &K, vector_type &S, matrix_type &Vt, matrix_type &U){
   //use 'S' for thin U,Vt matrices
   char jobu = 'S';
   char jobvt = 'S';

src/maxent.cpp

@@ -89,7 +89,7 @@ int main(int argc,const char** argv)
 
   std::string basename;
   if(parms.defaulted("BASENAME")){
-    basename = alps::fs::remove_extensions(parms.get_origin_name()) + ".out";
+    basename = alps::fs::remove_extensions(alps::origin_name(parms)) + ".out";
     parms["BASENAME"] = basename;
   }
   else
@@ -151,7 +151,7 @@ int main(int argc,const char** argv)
             ofstream_ spec_file;
             spec_file.open((basename+".varspec.dat").c_str());
             spec_file << "#omega mean_maxspec stdev_maxspec mean_avspec stdev_avspec" <<std::endl;
-            for (std::size_t  i=0; i<omega_grid.size(); ++i)
+            for (Eigen::Index i=0; i<omega_grid.size(); ++i)
               spec_file << omega_grid(i)
                 << " " << mean_max(i) << " " << stdev_max(i)
                 << " " << mean_av(i)  << " " << stdev_av(i) 
@@ -171,4 +171,3 @@ int main(int argc,const char** argv)
         std::cerr << "Caught Exception" << boost::current_exception_diagnostic_information();
     }
 }
-

src/maxent_backcont.cpp

@@ -39,12 +39,10 @@ vector_type Backcont::backcontinue(const vector_type &A){
 }
 double Backcont::max_error(const vector_type &y1, const vector_type &y2){
   double max_err = 0.0;
-#ifdef NDEBUG
   if(y1.size() != y2.size()){
     throw std::runtime_error("backcont vector size mismatch!");
   }
-#endif
-  for(int i=0;i<y1.size();i++){
+  for(Eigen::Index i=0;i<y1.size();i++){
     double delta = std::abs(y1(i)-y2(i));
     if(delta>max_err){
       max_err = delta;

src/maxent_backcont.hpp

@@ -11,7 +11,6 @@
 
 #pragma once
 #include "maxent.hpp"
-#include <alps/config.hpp> // needed to set up correct bindings
 
 ///This class contains necessities to analytically continue a 
 //real function to the imaginary axis