Clean up questionable code bits identified by clang sanitize. * data.cc (Fnorm): in %!test, don't test with a single value which exceeds range of single. * find.cc (find_nonzero_elem_idx): Remove unused variable assignment to n_to_find. * load-save.cc (Fsave): Declare argc just before usage. * ls-hdf5.cc (hdf5_read_next_data): Return from function immediately if type is unrecognized. * oct-stream.cc (octave_stream::read): Eliminate unnecessary initialization of elts_to_read. * octave-link.cc (F__octave_link_file_dialog__): Eliminate unnecessary initialization of nel. * pr-output.cc (set_real_format): Don't assign to function input variable digits. * str2double.cc (single_num): Initialize return value num at start of function. * ov-complex.cc (octave_complex::double_value, octave_complex::float_value): Eliminate unnecessary initialization of retval. * ov-flt-complex.cc (octave_float_complex::double_value, octave_float_complex::float_value): Eliminate unnecessary initialization of retval. * CSparse.cc (SparseComplexMatrix::inverse): Remove useless asignment to typ. * CSparse.cc (SparseComplexMatrix::bsolve): Initialize anorm before using in comparison. * MatrixType.cc (matrix_complex_probe): Initialize typ to MatrixType::Unknown. * dSparse.cc (SparseMatrix::inverse): Remove useless assignment to typ. * dSparse.cc (SparseMatrix::bsolve): Initialize anorm before using in comparison. * lo-specfun.cc (gammainc): Remove useless assignment to err. * data-conv.cc (oct_data_conv::string_to_data_type): Remove useless assignment to input_is_output. * str-vec.cc (string_vector::list_in_columns): Remove useless assignment to nc.

--- a/libinterp/octave-value/ov-complex.cc
+++ b/libinterp/octave-value/ov-complex.cc
@@ -105,7 +105,7 @@
 double
 octave_complex::double_value (bool force_conversion) const
 {
-  double retval = lo_ieee_nan_value ();
+  double retval;
 
   if (! force_conversion)
     gripe_implicit_conversion ("Octave:imag-to-real",
@@ -119,7 +119,7 @@
 float
 octave_complex::float_value (bool force_conversion) const
 {
-  float retval = lo_ieee_float_nan_value ();
+  float retval;
 
   if (! force_conversion)
     gripe_implicit_conversion ("Octave:imag-to-real",