NIntegrate Clears Notebook

Question

Bug introduced in 6 or 7 and resolved in 10.2

---

I have been doing some multi-dimensional numerical integration and noticing that when I try to evaluate certain integrals of functions I have defined, the evaluation stops almost instantly and clears all of my functions without producing an output. Below is a simplified version of the contents of my notebook.

ClearAll["Global`*"];

R3[x_?NumericQ, y_?NumericQ, n_?NumericQ] := Exp[-Abs[x - y]/n]*(x/n - y/n)^n;

b[T_?NumericQ] := 1/(T - I);

trans[x_?NumericQ] := x/(1 - x^2);

f[x_?NumericQ, y_?NumericQ, T_?NumericQ] :=  NIntegrate[R3[x, trans[z], 1]*R3[y, trans[z], 1]* Exp[I*b[T]*(x + trans[z])^2 - I*Conjugate[b[T]]*(y + trans[z])^2 + I*T]*D[trans[z], z], {z, 0, 1}, Method -> {"MonteCarlo", "MaxPoints" -> 10^2}];

g[T_?NumericQ] := NIntegrate[Abs[f[trans[x], trans[y], T]]^2*D[trans[x], x]*D[trans[y], y], {x, 0, 1}, {y, 0, 1}, Method -> {"MonteCarlo", "MaxPoints" -> 10^2}];

I have no problems evaluating the function f[x,y,T] at different points (I get the "failed to converge" error message, although it does not affect the evaluation), but when I try to evaluate, for example, g[0], the aforementioned problem of all function definitions getting cleared arises. I have never experienced this problem before and I'm wondering if I am making some silly mistake. Any help would be much appreciated.

Answer 1

The symptom of all symbols being cleared (turning from black to dark blue, then lighter blue on the next evaluation) indicates that the kernel crashed. While some kernel crashes can be induced by the user's unwary experimentation, in many cases it's just a bug. So it is here, as ilian notes in a comment:

This is a bug that has been fixed as of version 10.2.

In general, high-level functions like NIntegrate crashing the kernel is more likely to be a bug in those functions than due to anything the user has done, unless the input is somehow pathological.