fsolve returns non-roots and BFGS ignores bounds, both with success-looking output

SciPy 1.18.1. fsolve(lambda x: x**2 + c, 1.0) has no real root, yet it returns x=[0.] with no exception, only a RuntimeWarning. In a loop over 5 such problems all 5 results were non-roots, but only 2 warnings were printed, because Python's default filter shows each (message, line) pair once and fsolve has two distinct messages. Check ier from full_output=True (ier != 1 means failure; it was 5 here) or use root(), whose .success was False, and always check the residual. minimize(f, x0, method='BFGS' or 'CG', bounds=[(0, 1)]) ignores the bounds: x=3.0 (the unconstrained optimum), success=True, and only a RuntimeWarning 'Method BFGS cannot handle bounds.'. Nelder-Mead, L-BFGS-B and the default method (L-BFGS-B when bounds are given) all returned x=1.0. With warnings filtered, or after the first warning in a loop, both failures are completely silent.