Regression when compiled using Zig/Clang. test_diff_[correctout.txt]

[JamesParrott]

inf: fwd= appears before WARNING

E               AssertionError: Error on line num i: 37.  This line (and up to the 9 previous ones):
E                 Expected: 'hl/FULLhl: fwd=1 bwd=1'
E                 'one+two: fwd=3 bwd=3 TESTED NONLINEAR'
E                 "ERROR:  Data field 'nonexistantvariable' not found on net."
E                 'nonexistantvariable: did not compile'
E                 'ang: fwd=113.093 bwd=113.093'
E                 'euc: fwd=5.01976 bwd=5.01976'
E                 'hg: fwd=1 bwd=2'
E                 'hl: fwd=2 bwd=1'
E                 'hg+2.2*(hl+1.01): fwd=7.622 bwd=6.422 TESTED NONLINEAR'
E                 'WARNING: Formula evaluation gave infinite result for link 0',
E
E                  Actual: 'hl/FULLhl: fwd=1 bwd=1'
E                 'one+two: fwd=3 bwd=3 TESTED NONLINEAR'
E                 "ERROR:  Data field 'nonexistantvariable' not found on net."
E                 'nonexistantvariable: did not compile'
E                 'ang: fwd=113.093 bwd=113.093'
E                 'euc: fwd=5.01976 bwd=5.01976'
E                 'hg: fwd=1 bwd=2'
E                 'hl: fwd=2 bwd=1'
E                 'hg+2.2*(hl+1.01): fwd=7.622 bwd=6.422 TESTED NONLINEAR'
E                 'inf: fwd=WARNING: Formula evaluation gave infinite result for link 0'
E               assert False
E                +  where False = almost_equal('inf: fwd=WARNING: Formula evaluation gave infinite result for link 0', 'WARNING: Formula evaluation gave infinite result for link 0')

[JamesParrott]

Occurs on Linux too. inf: fwd= forms part of a subsequent line of correctout.txt. Both are in tests run with python -u though (in a subprocess from Pytest) so I don't think it's a buffering issue.