tests: fix gtest expect/assert if used incorrecrtly

[fnrizzi]

@mzuzek you raise a good point:

• we should explore all ops, ode, rom, tests to make sure that we use the correct gtest macro for accumulation vs exact equality
• one PR for each component

[mzuzek]

@fnrizzi

Here's the review of EXPECT_DOUBLE_EQ suspected usage in tests on develop at 7d2dbb9. Let me know what you think of categories and what's your opinion on which cases should be fixed (i.e. exact EXPECT_DOUBLE_EQ replaced with tolerance-based EXPECT_NEAR, like in #438).

literal

Comparisons against literal expressions - like EXPECT_DOUBLE_EQ(R(7), 21*1.+22*2.+23*3.+1.) - were excluded as not suspected (i.e. correctly assuming highest level of accuracy). This includes also indirect cases where a literal is first assigned to a variable and that variable is then used in comparison (for instance see _ops/ops_eigendiag.cc:244-252).

value

These cases compare value obtained from tested calculation against value obtained from reference calculation. Most probably should use tolerance-based comparisons instead of exact ones (even if integer values are actually used at this particular moment - like for the time steps).

• _odesteppers:
  • _ode_steppers/ode_rk4_custom_ind_var_typeeigen.cc:14-16
  • _ode_steppers/ode_rk4_simple_correctnesseigen.cc:12-14
  • _ode_steppers/ode_bdf1_simple_correctnesseigen.cc:61-63
  • _ode_steppers/ode_bdf2_simple_correctnesseigen.cc: lines 31-33 and 62-64
  • _ode_steppers/ode_all_explicit_schemes_fixed_mass_matrix_correctnesseigen.cc:89
  • _ode_steppers/ode_all_explicit_schemes_varying_mass_matrix_correctnesseigen.cc: lines 91 and 113
• rom:
  • _rom/lspgunsteady/main1.cc: lines 116, 122, 149, 155, 185, 191, 218, 224
  • _rom/lspgunsteady/main4.cc: lines 70, 76, 103, 109, 139, 145, 172, 178
  • _rom/lspgsteady/main1.cc: lines 89, 113
  • _rom/lspgsteady/main3.cc: lines 96, 120
  • _rom/lspgsteady/main4.cc: lines 98, 122

    literal sqrt()

These cases compare against std::sqrt(NUM_LITERAL) - assuming certain accuracy of the standard function (correctly or not):

• ops
  • _ops/ops_epetravector.cc:93
  • _ops/ops_tpetravector.cc:99
  • _ops/ops_tpetra_blockvector.cc:99
• _solversnonlinear:
  • _solversnonlinear/operators.cc: lines: 31, 60, 249, 268, 297, 319, 352, 429

[fnrizzi]

i agree with this , if you want to make a PR for the changes that would be great