Stan Segfault

[David Manheim]

Segfault happened. I'm unsure if this is enough information to help anyone figure out what occurred, but I figured I'd at least mention that it happened, in case there was some chance it was helpful.

Cheers,

David

[Ben Goodrich]

On Tuesday, March 14, 2017 at 9:14:42 AM UTC-4, David Manheim wrote:

Segfault happened. I'm unsure if this is enough information to help anyone figure out what occurred, but I figured I'd at least mention that it happened, in case there was some chance it was helpful.

This tends to happen when Rcpp is compiled with a different compiler or compiler setting than the Stan program.

Ben

[David Manheim]

I'm running Rcpp 12.7, and the latest, released in January, is 12.9.

I'm unsure how to get Rcpp 12.8 installed on this server, which I assume was used given that Rstan 2.14.1 was released in December.

I may not be able to recompile stan on the server I am using due to permissions, and I'm unsure I can get them to install a previous version of Rcpp either :(

Thanks,

David

[Ben Goodrich]

On Friday, March 17, 2017 at 10:50:23 AM UTC-4, David Manheim wrote:

It happened again - so how can I fix it?

I may not be able to recompile stan on the server I am using due to permissions; is there a different approach? Can I perhaps use a different version of Rcpp - and if so, which version does Rstan use?

It uses whatever version of Rcpp is on the server. But with Ubuntu servers, sometimes core packages are installed as binaries from the official repositories, i.e. r-cran-rcpp, in which case Rcpp was compiled with whatever setting were used in the repositories. If you have StanHeaders, rstan, etc. installed in your home directory on the server, it would usually be possible to install Rcpp there as well, in which case it would use the version in your home directory rather than the global one.

Ben

[David Manheim]

That's very strange - I don't have Rcpp or rstan installed under my user account, both are the installed on the server. And it's running Redhat.

The error occurred on the 4th of 6 runs that were being run by a loop in R, so it's not always happening; is it possible that this is a different issue?

[Ben Goodrich]

On Friday, March 17, 2017 at 4:33:29 PM UTC-4, David Manheim wrote:

That's very strange - I don't have Rcpp or rstan installed under my user account, both are the installed on the server. And it's running Redhat.

The error occurred on the 4th of 6 runs that were being run by a loop in R, so it's not always happening; is it possible that this is a different issue?

Does this

https://github.com/stan-dev/rstan/issues/399

sound familiar?

[David Manheim]

I'm not extracting anything, so that shouldn't be it - I save each model independently to a file, then do the next one.

[Ben Goodrich]

On Monday, March 20, 2017 at 11:29:51 AM UTC-4, David Manheim wrote:

I'm not extracting anything, so that shouldn't be it - I save each model independently to a file, then do the next one.

There is a new rstan on CRAN now, so see if you can get your system administrator to install it and maybe it will fix your problem.

Ben