This is not an answer but I have tried to reproduce this bug with a simpler example and I obtained this:
u(x) = sqrt((1-tan(x)^2)) f(x) = pi/2-arccos(u(x)); r(x) = f(x)*cos(x) integral(r(x), (x, 0, pi/4)) 1.11072073453959 But numerical_integral(r(x), 0, pi/4) (0.785398166410623, 6.110348980000323e-07) (Again an error of a factor sqrt(2)). Strangely if you change f(x) into simply arccos(x) instead of pi/2-arccos(x), a test seems to be peformed. A warning is issued and no answer is given (the output after the warning is again integrate(arccos(sqrt(-tan(x)^2 + 1))*cos(x), x, 0, 1/4*pi) ) u(x) = sqrt((1-tan(x)^2)) f(x) = arccos(u(x)); r(x) = f(x)*cos(x) integral(r(x), (x, 0, pi/4)) Warning, integration of abs or sign assumes constant sign by intervals (correct if the argument is real): Check [abs(t_nostep)] Warning, integration of abs or sign assumes constant sign by intervals (correct if the argument is real): Check [abs(t_nostep^2-1)] Warning, integration of abs or sign assumes constant sign by intervals (correct if the argument is real): Check [abs(t_nostep^2-1)] Warning, choosing root of [1,0,%%%{4,[2,4]%%%}+%%%{-6,[2,2]%%%}+%%%{2,[2,0]%%%}+%%%{-6,[0,4]%%%}+%%%{10,[0,2]%%%}+%%%{-4,[0,0]%%%},0,%%%{4,[4,8]%%%}+%%%{-12,[4,6]%%%}+%%%{13,[4,4]%%%}+%%%{-6,[4,2]%%%}+%%%{1,[4,0]%%%}+%%%{4,[2,8]%%%}+%%%{-10,[2,6]%%%}+%%%{8,[2,4]%%%}+%%%{-2,[2,2]%%%}+%%%{1,[0,8]%%%}+%%%{-2,[0,6]%%%}+%%%{1,[0,4]%%%}] at parameters values [49,-6] Warning, choosing root of [1,0,%%%{4,[2,4]%%%}+%%%{-6,[2,2]%%%}+%%%{2,[2,0]%%%}+%%%{-6,[0,4]%%%}+%%%{10,[0,2]%%%}+%%%{-4,[0,0]%%%},0,%%%{4,[4,8]%%%}+%%%{-12,[4,6]%%%}+%%%{13,[4,4]%%%}+%%%{-6,[4,2]%%%}+%%%{1,[4,0]%%%}+%%%{4,[2,8]%%%}+%%%{-10,[2,6]%%%}+%%%{8,[2,4]%%%}+%%%{-2,[2,2]%%%}+%%%{1,[0,8]%%%}+%%%{-2,[0,6]%%%}+%%%{1,[0,4]%%%}] at parameters values [0,81] Discontinuities at zeroes of t_nostep^2-1 were not checked Discontinuities at zeroes of t_nostep^2-1 were not checked Warning, integration of abs or sign assumes constant sign by intervals (correct if the argument is real): Check [abs(t_nostep)] Error while checking exact value with approximate value, returning both! integrate(arccos(sqrt(-tan(x)^2 + 1))*cos(x), x, 0, 1/4*pi) Emmanuel El dom, 30 abr 2023 a las 9:35, Pong (<wypon...@gmail.com>) escribió: > The codes > > x,y = var('x,y'); > f(x) = acos(sqrt((1-tan(x)^2)/2)); > g(x) = integral(sin(y)^4,(y,f(x),pi-f(x))); > h(x) = sin(x)^2*cos(x)*g(x); > integral(h(x),(x,-pi/4,pi/4)), numerical_integral(h(x),-pi/4,pi/4) > > produce > > (1/16*sqrt(2)*pi, (0.1963495451106892, 9.705160370278192e-07)) > > SageMath version: 9.8 on Ubuntu 22.04 (SAGE was complied from source) > > We believe the numerical answer is correct (that should be > pi/16=0.1963....) since we got that answer by computing the integral in > another way by hand. > > We were surprised that 'integral' can give us an answer and even more > surprised by the fact that it is off by a factor of sqrt(2) from the answer > given by 'numerical_integral'. > > Any insight of what's happening here? > > --Pong > > -- > You received this message because you are subscribed to the Google Groups > "sage-devel" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to sage-devel+unsubscr...@googlegroups.com. > To view this discussion on the web visit > https://groups.google.com/d/msgid/sage-devel/f1b3157d-20b5-4e7f-aa72-8046f84d5183n%40googlegroups.com > <https://groups.google.com/d/msgid/sage-devel/f1b3157d-20b5-4e7f-aa72-8046f84d5183n%40googlegroups.com?utm_medium=email&utm_source=footer> > . > -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/CAC2RnLpgss%3DW7HtD%3Dg%3DWZCi27EZPiexJ9ixr9WGU1gDVCdeZ%3Dg%40mail.gmail.com.