Eigen out of bounds access error with simple model

[bbbales2]

Here is the model:

transformed data {
  vector[1] y = to_vector({ 1.0 });
}

parameters {
  real mu;
}

model {
  y[0] ~ normal(mu, 1.0);
}

If I build and run it:

./model sample

I get an Eigen out of bounds error and the program stops:

model: stan/lib/stan_math/lib/eigen_3.3.3/Eigen/src/Core/DenseCoeffsBase.h:162: Eigen::DenseCoeffsBase<type-parameter-0-0, 0>::CoeffReturnType Eigen::DenseCoeffsBase<Eigen::Matrix<double, -1, 1, 0, -1, 1>, 0>::operator[](Eigen::Index) const [Derived = Eigen::Matrix<double, -1, 1, 0, -1, 1>, Level = 0]: Assertion `index >= 0 && index < size()' failed.
Aborted (core dumped)

There should be some sort of Stan out of bounds access error I'm pretty sure. This is from cmdstan in develop. I assume this is a compiler thing, but maybe it is math somewhere. Can move the issue if that is the case.

[bob-carpenter]

I think this may be related to the expression creating the vector. If you create an array without that expression logic, is there still a problem?

[bbbales2]

Same error with:

transformed data {
  vector[1] y;
  y[1] = 1.0;
}
...

And the error goes away with the statement:

y[1] ~ normal(mu, 1.0);

[bob-carpenter]

Ah, I missed that the y[0] was indexing out of range. So definitely an error checking issue, not a construction of vectors issue. We need to catch the bad indexing to report an error.