Compile error when calling the function stan_model in an R Jupyter notebook

Summary:

I get a compile error when calling the function stan_model in an R Jupyter notebook. When running the same code from the terminal no error occurs.

Description:

I'm trying to run the following toy model in rstan:

# In R:
# Load necessary libraries and set up multi-core processing for Stan
options(warn=-1, message =-1)
library(dplyr); library(ggplot2); library(rstan); library(reshape2)
rstan_options(auto_write = TRUE)
options(mc.cores = parallel::detectCores())

# In R, or you could save the contents of the string in a file with .stan file type

dgp_string <- "

functions {
    /**
   * Return draws from a linear regression with data matrix X,
   * coefficients beta, and student-t noise with degrees of freedom nu
   * and scale sigma.
   *
   * @param X Data matrix (N x P)
   * @param beta Coefficient vector (P x 1)
   * @param nu Residual distribution degrees of freedom.
   * @param sigma Residual distribution scale.
   * @return Return an N-vector of draws from the model.
   */

    vector dgp_rng(matrix X, vector beta, real nu, real sigma) {
      vector[rows(X)] y; // define the output vector to be as long as the number of rows in X
      
      // Now fill it in
      for (n in 1:rows(X))
        y[n] <- student_t_rng(nu, X[n] * beta, sigma);
      return y;
   }
}
data {
 // If we were estimating a model, we'd define the data inputs here
}
parameters {
  // ... and the parameters we want to estimate would go in here
}
model {
  // This is where the probability model we want to estimate would go
}
"

# Generate a matrix of random numbers, and values for beta, nu and sigma

set.seed(42) # Set the random number generator seed so that we get the same parameters
N <- 1000 # Number of observations
P <- 10 # Number of covariates
X <- matrix(rnorm(N*P), N, P) # generate an N*P covariate matrix of random data
nu <- 5 # Set degrees of freedom
sigma <- 5 # And scale parameter
beta <- rnorm(10) # Generate some random coefficients that we'll try to recover
# Make sure the first element of beta is positive as in our chosen DGP
beta[1] <- abs(beta[1])

# Compile the script
compiled_function <- stan_model(model_code = dgp_string) # you could use file = "path/to/yourfile.stan" if you have saved it as so

Current Output:

If applicable, any relevant output from RStan.

I get the following error message:

(...)
Error in compileCode(f, code, language = language, verbose = verbose): Compilation ERROR, function(s)/method(s) not created! **********************************************************************
* Please run brazil-build rather than make.  Your build might not work
* correctly without running brazil-build.
**********************************************************************
cc1plus: error: unrecognized command line option "-mbmi-mavx"
make: *** [file3bf5713249a.o] Error 1
Traceback:

1. stan_model(model_code = dgp_string)
2. cxxfunctionplus(signature(), body = paste(" return Rcpp::wrap(\"", 
 .     model_name, "\");", sep = ""), includes = inc, plugin = "rstan", 
 .     save_dso = save_dso | auto_write, module_name = paste("stan_fit4", 
 .         model_cppname, "_mod", sep = ""), verbose = verbose)
3. cxxfunction(sig = sig, body = body, plugin = plugin, includes = includes, 
 .     settings = settings, ..., verbose = verbose)
4. compileCode(f, code, language = language, verbose = verbose)
5. stop(paste("Compilation ERROR, function(s)/method(s) not created!", 
 .     paste(errmsg, collapse = "\n")))

RStan and R version:

sessionInfo()

R version 3.3.2 (2016-10-31)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Amazon Linux Bare Metal release 2012.03

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] reshape2_1.4.2       rstan_2.12.1         StanHeaders_2.12.0-1
[4] ggplot2_2.2.0        dplyr_0.5.0         

loaded via a namespace (and not attached):
 [1] Rcpp_0.12.8      magrittr_1.5     munsell_0.4.3    uuid_0.1-2      
 [5] colorspace_1.3-1 R6_2.2.0         stringr_1.1.0    plyr_1.8.4      
 [9] tools_3.3.2      parallel_3.3.2   grid_3.3.2       gtable_0.2.0    
[13] DBI_0.5-1        lazyeval_0.2.0   assertthat_0.1   digest_0.6.10   
[17] tibble_1.2       crayon_1.3.2     gridExtra_2.2.1  IRdisplay_0.4.4 
[21] repr_0.10        IRkernel_0.7.1   inline_0.3.14    evaluate_0.10   
[25] pbdZMQ_0.2-4     stringi_1.1.2    scales_0.4.1     stats4_3.3.2    
[29] jsonlite_1.1    

Operating System:

cat /etc/redhat-release

Red Hat Enterprise Linux Server release 5.3 (Tikanga)

Start by reproducing the stan_model call in the R Jupyter notebook and from the terminal, using the reported RStan versions and compiler error for -mbmi-mavx. Inspect the compileCode path and RStan build configuration; done means the notebook compilation succeeds without the unrecognized compiler option, or the incompatibility is clearly documented.