Dear maintainer:

During a rebuild of all packages in unstable, this package failed to build.

Below you will find the last part of the build log (probably the most
relevant part, but not necessarily). If required, the full build log
is available here:

https://people.debian.org/~sanvila/build-logs/202608/

About the archive rebuild: The build was made on virtual machines from AWS,
using sbuild and a reduced chroot with only build-essential packages.

If you cannot reproduce the bug please contact me privately, as I
am willing to provide ssh access to a virtual machine where the bug is
fully reproducible.

If this is really a bug in one of the build-depends, please use
reassign and add an affects on src:octave-statistics, so that this is still
visible in the BTS web page for this package.

Thanks.
--------------------------------------------------------------------------------
[...]
 Xs = [randn(30,2)*0.4 + 2; randn(30,2)*0.4 - 2];
 Ys = [ones(30,1); 2*ones(30,1)];
 M = fcnntrain (Xs, Ys, 8, "relu", "softmax", 1, 0.05, 60, false, 1);
 assert_equal (numel (M.Loss), 60);
 assert_equal (numel (M.Accuracy), 60);
 assert_equal (any (M.Loss != 0), true);
 assert_equal (M.Loss(end) < M.Loss(1), true);
 assert_equal (M.Accuracy(end) >= M.Accuracy(1), true);
***** error <fcnntrain: too many input arguments.> ...
 fcnntrain (X, Y, 10, "sigmoid", "sigmoid", 1, 0.025, 50, false, 0, struct (), 0);
***** error <fcnntrain: 'SolverOptions' must be a scalar struct.> ...
 fcnntrain (X, Y, 10, "sigmoid", "sigmoid", 1, 0.025, 50, false, 0, 0);
***** error <fcnntrain: 'Solver' must be 'sgd' or 'lbfgs'.> ...
 fcnntrain (X, Y, 10, "sigmoid", "sigmoid", 1, 0.025, 50, false, 0, ...
            struct ("Solver", "bogus"));
***** error <fcnntrain: 'LossFunction' must be a numeric scalar value.> ...
 fcnntrain (X, Y, 10, "sigmoid", "sigmoid", 1, 0.025, 50, false, 'ce');
***** error <fcnntrain: 'LossFunction' must be a numeric scalar value.> ...
 fcnntrain (X, Y, 10, "sigmoid", "sigmoid", 1, 0.025, 50, false, [0, 1]);
***** error <fcnntrain: invalid 'LossFunction' code.> ...
 fcnntrain (X, Y, 10, "sigmoid", "sigmoid", 1, 0.025, 50, false, 3);
***** error <fcnntrain: invalid 'LossFunction' code.> ...
 fcnntrain (X, Y, 10, "sigmoid", "sigmoid", 1, 0.025, 50, false, -1);
***** test
 rand ('seed', 42);
 randn ('seed', 42);
 Xr = linspace (-2, 2, 60)';
 Yr = 3 * Xr - 1;
 M = fcnntrain (Xr, Yr, [8, 8], "relu", "linear", 1, 0.005, 300, false, 2);
 assert_equal (fieldnames (M), {'LayerWeights'; 'Activations'; 'Loss'});
 assert_equal (rows (M.LayerWeights{end}), 1);
 assert_equal (M.Loss(end) < M.Loss(1), true);
***** test
 rand ('seed', 42);
 Xr = linspace (0, 1, 40)';
 Yr = 5 * Xr + 2;
 M = fcnntrain (Xr, Yr, 10, "relu", "linear", 1, 0.005, 200, false, 2);
 Wm = cellfun (@(m) m(:,1:end-1), M.LayerWeights, "UniformOutput", false);
 Bm = cellfun (@(m) m(:,end), M.LayerWeights, "UniformOutput", false);
 [~, yFit] = fcnnpredict (Wm, Bm, "relu", "linear", Xr);
 assert_equal (M.Loss(end), mean ((Yr - yFit) .^ 2), 1e-12);
***** test
 rand ('seed', 42);
 Xr = linspace (0, 1, 30)';
 M = fcnntrain (Xr, [Xr, 2 * Xr, 3 * Xr], 6, "relu", "linear", 1, 0.005, ...
                50, false, 2);
 assert_equal (rows (M.LayerWeights{end}), 3);
 Wm = cellfun (@(m) m(:,1:end-1), M.LayerWeights, "UniformOutput", false);
 Bm = cellfun (@(m) m(:,end), M.LayerWeights, "UniformOutput", false);
 [~, yFit] = fcnnpredict (Wm, Bm, "relu", "linear", Xr);
 assert_equal (columns (yFit), 3);
***** test
 rand ('seed', 42);
 Xr = linspace (0, 1, 20)';
 Yr = linspace (-3.5, 2.25, 20)';
 M = fcnntrain (Xr, Yr, 6, "relu", "linear", 1, 0.005, 50, false, 2);
 assert_equal (all (isfinite (M.Loss)), true);
***** error <fcnntrain: Y must be finite.> ...
 fcnntrain ([1; 2; 3], [1; Inf; 3], 4, "relu", "linear", 1, 0.01, 10, false, 2);
***** error <fcnntrain: Y must be finite.> ...
 fcnntrain ([1; 2; 3], [1; NaN; 3], 4, "relu", "linear", 1, 0.01, 10, false, 2);
***** test
 so = struct ("Solver", "lbfgs");
 M = fcnntrain (X, Y, 10, "relu", "softmax", 1, 0.005, 100, false, 1, so);
 assert_equal (fieldnames (M), ...
               {'LayerWeights'; 'Activations'; 'Loss'; 'Gradient'; ...
                'Step'; 'Criterion'});
 assert_equal (M.Loss(end) < M.Loss(1), true);
 assert_equal (numel (M.Gradient), numel (M.Loss));
 assert_equal (numel (M.Step), numel (M.Loss));
***** test
 rand ("state", 3); randn ("state", 3);
 Ms = fcnntrain (X, Y, 10, "relu", "softmax", 1, 0.005, 200, false, 1);
 rand ("state", 3); randn ("state", 3);
 so = struct ("Solver", "lbfgs");
 Ml = fcnntrain (X, Y, 10, "relu", "softmax", 1, 0.005, 200, false, 1, so);
 assert_equal (Ml.Loss(end) < Ms.Loss(end), true);
 assert_equal (numel (Ml.Loss) < numel (Ms.Loss), true);
***** test
 rand ("state", 5); randn ("state", 5);
 Ma = fcnntrain (X, Y, 10, "relu", "softmax", 1, 0.005, 30, false, 1);
 rand ("state", 5); randn ("state", 5);
 so = struct ("Solver", "sgd");
 Mb = fcnntrain (X, Y, 10, "relu", "softmax", 1, 0.005, 30, false, 1, so);
 assert_equal (Mb.Loss, Ma.Loss);
 assert_equal (Mb.LayerWeights, Ma.LayerWeights);
***** test
 rand ("state", 9); randn ("state", 9);
 so = struct ("Solver", "lbfgs", "GradientTolerance", 1e3);
 Ma = fcnntrain (X, Y, 10, "relu", "softmax", 1, 0.005, 100, false, 1, so);
 rand ("state", 9); randn ("state", 9);
 so = struct ("Solver", "lbfgs", "GradientTolerance", 1e-8);
 Mb = fcnntrain (X, Y, 10, "relu", "softmax", 1, 0.005, 100, false, 1, so);
 assert_equal (Ma.Criterion, "Relative gradient tolerance reached.");
 assert_equal (numel (Ma.Loss) < numel (Mb.Loss), true);
51 tests, 51 passed, 0 known failure, 0 skipped
[src/editDistance.cc]
***** error <editDistance: too many input arguments.> d = editDistance (1, 2, 3, 4);
***** error <editDistance: too many output arguments.> ...
 [C, IA, IC, I] = editDistance ({"AS","SD","AD"}, 1);
***** error <editDistance: too many output arguments.> ...
 [C, IA] = editDistance ({"AS","SD","AD"});
***** error <editDistance: minDist must be a scalar value.> ...
 d = editDistance ({"AS","SD","AD"}, [1, 2]);
***** error <editDistance: minDist must be a nonnegative integer.> ...
 d = editDistance ({"AS","SD","AD"}, -2);
***** error <editDistance: minDist must be a nonnegative integer.> ...
 d = editDistance ({"AS","SD","AD"}, 1.25);
***** error <editDistance: minDist must be a scalar value.> ...
 d = editDistance ({"AS","SD","AD"}, {"AS","SD","AD"}, [1, 2]);
***** error <editDistance: minDist must be a nonnegative integer.> ...
 d = editDistance ({"AS","SD","AD"}, {"AS","SD","AD"}, -2);
***** error <editDistance: minDist must be a nonnegative integer.> ...
 d = editDistance ({"AS","SD","AD"}, {"AS","SD","AD"}, 1.25);
***** error <editDistance: minDist must be a scalar value.> ...
 d = editDistance ("string1", "string2", [1, 2]);
***** error <editDistance: minDist must be a nonnegative integer.> ...
 d = editDistance ("string1", "string2", -2);
***** error <editDistance: minDist must be a nonnegative integer.> ...
 d = editDistance ("string1", "string2", 1.25);
***** error <editDistance: tokenizedDocument must contain cellstr arrays.> ...
 d = editDistance ({{"string1", "string2"}, 2});
***** error <editDistance: tokenizedDocument must contain cellstr arrays.> ...
 d = editDistance ({{"string1", "string2"}, 2}, 2);
***** error <editDistance: STR1 must be a cellstr.> ...
 d = editDistance ([1, 2, 3]);
***** error <editDistance: STR1 must be a cellstr.> ...
 d = editDistance (["AS","SD","AD","AS"]);
***** error <editDistance: STR1 must be a cellstr.> ...
 d = editDistance (["AS","SD","AD"], 2);
***** error <editDistance: STR1 and STR2 must be either strings or cellstr.> ...
 d = editDistance (logical ([1,2,3]), {"AS","AS","AD"});
***** error <editDistance: STR1 and STR2 must be either strings or cellstr.> ...
 d = editDistance ({"AS","SD","AD"}, logical ([1,2,3]));
***** error <editDistance: STR1 and STR2 must be either strings or cellstr.> ...
 d = editDistance ([1,2,3], {"AS","AS","AD"});
***** error <editDistance: first tokenizedDocument does not contain cellstr arrays.> ...
 d = editDistance ({1,2,3}, {"AS","SD","AD"});
***** error <editDistance: second tokenizedDocument does not contain cellstr arrays.> ...
 d = editDistance ({"AS","SD","AD"}, {1,2,3});
***** error <editDistance: cellstr input arguments size mismatch.> ...
 d = editDistance ({"AS","SD","AD"}, {"AS", "AS"});
***** test
 d = editDistance ({"AS","SD","AD"});
 assert_equal (d, [2; 1; 1]);
 assert_equal (class (d), "double");
***** test
 C = editDistance ({"AS","SD","AD"}, 1);
 assert_equal (iscellstr (C), true);
 assert_equal (C, {"AS";"SD"});
***** test
 [C, IA] = editDistance ({"AS","SD","AD"}, 1);
 assert_equal (class (IA), "double");
 assert_equal (IA, [1;2]);
***** test
 A = {"ASS"; "SDS"; "FDE"; "EDS"; "OPA"};
 [C, IA] = editDistance (A, 2, "OutputAllIndices", false);
 assert_equal (class (IA), "double");
 assert_equal (A(IA), C);
***** test
 A = {"ASS"; "SDS"; "FDE"; "EDS"; "OPA"};
 [C, IA] = editDistance (A, 2, "OutputAllIndices", true);
 assert_equal (class (IA), "cell");
 assert_equal (C, {"ASS"; "FDE"; "OPA"});
 assert_equal (A(IA{1}), {"ASS"; "SDS"; "EDS"});
 assert_equal (A(IA{2}), {"FDE"; "EDS"});
 assert_equal (A(IA{3}), {"OPA"});
***** test
 A = {"ASS"; "SDS"; "FDE"; "EDS"; "OPA"};
 [C, IA, IC] = editDistance (A, 2);
 assert_equal (class (IA), "double");
 assert_equal (A(IA), C);
 assert_equal (IC, [1; 1; 3; 1; 5]);
***** test
 d = editDistance ({"AS","SD","AD"}, {"AS", "AD", "SE"});
 assert_equal (d, [0; 1; 2]);
 assert_equal (class (d), "double");
***** test
 d = editDistance ({"AS","SD","AD"}, {"AS"});
 assert_equal (d, [0; 2; 1]);
 assert_equal (class (d), "double");
***** test
 d = editDistance ({"AS"}, {"AS","SD","AD"});
 assert_equal (d, [0; 2; 1]);
 assert_equal (class (d), "double");
***** test
 b = editDistance ("Octave", "octave");
 assert_equal (b, 1);
 assert_equal (class (b), "double");
33 tests, 33 passed, 0 known failure, 0 skipped
[src/svmpredict.cc]
***** test
 # Test 1: Standard C-SVC Prediction (Original Regression Test)
 [L, D] = libsvmread (file_in_loadpath ("heart_scale.dat"));
 model = svmtrain (L, D, '-c 1 -g 0.07');
 [predict_label, accuracy, dec_values] = svmpredict (L, D, model);
 assert_equal (size (predict_label), size (dec_values));
 assert_equal (accuracy, [86.666, 0.533, 0.533]', [1e-3, 1e-3, 1e-3]');
 assert_equal (dec_values(1), 1.225836001973273, 1e-14);
 assert_equal (dec_values(2), -0.3212992933043805, 1e-14);
 assert_equal (predict_label(1), 1);

***** test
 # A single testing instance used to write through a freed pointer, since
 # Octave stores a 1x1 result as a scalar and matrix_value() then returns a
 # temporary.  Every single-row query must match the batch answer.
 [L, D] = libsvmread (file_in_loadpath ("heart_scale.dat"));
 model = svmtrain (L, D, '-c 1 -g 0.07');
 [bl, ~, bd] = svmpredict (L, D, model);
 for i = [1, 2, 7, 130, numel(L)]
   [l, ~, d] = svmpredict (L(i), D(i,:), model);
   assert_equal (size (l), [1, 1]);
   assert_equal (l, bl(i));
   assert_equal (d, bd(i), 1e-12);
 endfor

***** test
 # Test 2: One-Class Probability (NEW LIBSVM 3.36 FEATURE)
 [L, D] = libsvmread (file_in_loadpath ("heart_scale.dat"));
 # Train One-Class (-s 2) with Probability (-b 1)
 model_oc = svmtrain (L, D, '-s 2 -n 0.1 -g 0.07 -b 1');
 assert_equal (isstruct (model_oc), true);
 # Predict with Probability (-b 1)
 [pred, acc, probs] = svmpredict (L, D, model_oc, '-b 1');

 # Detail Check A: Output must be N x 2 (Column 1: Normal, Column 2: Outlier)
 assert_equal (size (probs), [length(L), 2]);

 # Detail Check B: Probabilities must sum to 1.0 for every instance
 assert_equal (sum (probs, 2), ones (length(L), 1), 1e-5);

 # Detail Check C: Values must be valid probabilities [0, 1]
 assert_equal (all (probs >= 0 & probs <= 1, 'all'), true);
 clear model_oc

***** test
 # Test 3: One-Class Decision Values (Standard Check)
 # Verifies that the upgrade didn't break standard One-Class prediction (-b 0)
 [L, D] = libsvmread (file_in_loadpath ("heart_scale.dat"));
 model_oc = svmtrain (L, D, '-s 2 -n 0.1 -g 0.07');
 [pred, acc, dec] = svmpredict (L, D, model_oc);
 # Standard One-Class output is N x 1 (Scalar decision values)
 assert_equal (size (dec), [length(L), 1]);
 clear model_oc

***** shared L, D, model
 # Test 4: Error Handling (Original Checks)
 [L, D] = libsvmread (file_in_loadpath ("heart_scale.dat"));
 model = svmtrain (L, D, '-c 1 -g 0.07');

***** error <svmpredict: wrong number of output arguments.> ...
 [p, a] = svmpredict (L, D, model);
***** error <svmpredict: wrong number of input arguments.> p = svmpredict (L, D);
***** error <svmpredict: label vector and instance matrix must be double.> ...
 p = svmpredict (single (L), D, model);
***** error <svmpredict: model should be a struct array.> p = svmpredict (L, D, 123);
8 tests, 8 passed, 0 known failure, 0 skipped
Done running the unit tests.
Summary: 16761 tests, 16758 passed, 0 known failures, 0 skipped
Some tests failed.  Giving up...
make: *** [debian/rules:5: binary] Error 1
dpkg-buildpackage: error: debian/rules binary subprocess failed with exit status 2
--------------------------------------------------------------------------------