An updated version 0.2.2 of the random-number generator tester RDieHarder (based on the DieHarder suite developed / maintained by Robert Brown with contributions by David Bauer and myself) is now on CRAN.
I should dub this the ‘due to Brian Ripley’ release. He sent me a detailed five-point email a few days ago which detailed a change I could not have tested (“no access”), a change I would not have known (“somewhat obscure C language bit-level manipulation”), a change I had missed (how my build setup failed for M1mac), another advanced C level fix, and one more simple fix I actually knew. Speechless. The man (I presume) does not sleep and is just so generous with his time and expertise.
So based on the input I rejigged the package over the weekend and made two more (substantial) changes. First, extending on what 0.2.0 brought, I will no longer attempt to use an external libdieharder library (or build one on the fly) – that was issue one. Now we just declare all C files as dependents of the package shared library, and things are simpler and more consistent. Sadly, that also implies “everything is in the package” so I had to edit out a metric ton of stdout or exit() reference with the appropriate R C API hooks to appease the CRAN Policy deities. Win some, loose some. But the package is now simpler, and cleaner, and should be in good standing. (Or so one hopes. Earlier today, and within hours of it hitting CRAN, I got an issue ticket from a motivated user about yet another (“mostly harmless” in the Douglas Adams sense) compiler warning… Good now too.)
Thanks to CRANberries, you can also look at the most recent diff.
If you like this or other open-source work I do, you can now sponsor me at GitHub.
This post by Dirk Eddelbuettel originated on his Thinking inside the box blog. Please report excessive re-aggregation in third-party for-profit settings.