argument testing alone isn't enough

[Daniel Lee]

The arguments were created consistently, just wrong. The argument test will have tested it the way it was built. It cast the double value to a character and moved on.

The seg fault happened with the cast of the pointer in CmdStan. It was cast to the wrong type. There isn't a way to fix that using pointers. It's programmer error and a test from within Stan wouldn't have caught it.

The real fix is to do this without pointers. I think we're heading that direction, but in the future.

Daniel

---------- Forwarded message ----------
From: Michael Betancourt <notifi...@github.com>
Date: Tue, Oct 27, 2015 at 6:00 PM
Subject: Re: [stan] Fixes variational eta argument (#1645)
To: stan-dev/stan <st...@noreply.github.com>
Cc: Daniel Lee <bea...@alum.mit.edu>


Create an issue — my best guess is that the seg fault is short circuiting
the test which causes it not to fail.