A challenging ODE question: find an integrating factor

[AdrianZ]

This is the ODE: y' + siny + xcosy + x = 0.
The problem is: Find an integrating factor for the ODE above.

You can see my solution to the ODE here: http://www.physicsforums.com/showthread.php?t=543662. from my solution it seems that e^x(sec^2(y/2)) must be an integrating factor. but I fail to show that it's an integrating factor.
I guess it must be truly challenging to find an integrating factor for it :(

[lurflurf]

I think that is an integrating factor, recall the general test
u is an integrating factor of N y'+M if

My-Nx=(ux/u)N-(uy/u)M
where subscripts are partial derivatives and for this equation we have
M=sin(y) + x cos(y) + x
N=1
u=exsec2(y/2)=2ex/(1+cos(y))
you will need to use some simple trigonometric identities

[AdrianZ]

I think that is an integrating factor, recall the general test
u is an integrating factor of N y'+M if

My-Nx=(ux/u)N-(uy/u)M
where subscripts are partial derivatives and for this equation we have
M=sin(y) + x cos(y) + x
N=1
u=exsec2(y/2)
you will need to use some trigonometric identities for half angles

What he says doesn't make sense to me to be honest. I've solved the ODE, and I'm 99% sure that exsec2(y/2) is an integrating factor. he tells me that I should multiply the integrating factor and show that it turns the ODE into an exact differential because his problem asks me to find an integrating factor for it, not solving the ODE.

[lurflurf]

Yes if My-Nx=(ux/u)N-(uy/u)M
then the differential is exact this follows from requiring
(uM)y-(uN)x=0
then it is possible to choose a P such that
N=Py/u
M=Px/u
hence
(P)'/u=N y'+M
as desired