fix checksums for pkgmaker/rngtools/RWeka/RcppProgress extensions in R 3.4.4 easyconfigs

[boegel]

No idea what happened, but when downloading the following files from CRAN now, the checksums are different than before.

That's strange, especially for pkgmaker_0.22.tar.gz and rngtools_1.2.4.tar.gz, both of which were downloaded on Jan 16th 2015 and haven't changed since (until now, that is).

[boegel]

I checked the differences between the old and the new tarballs:

• pkgmaker_0.22.tar.gz: no changes according to diff -ru
• rngtools_1.2.4.tar.gz: no changes according to diff -ru
• RcppProgress_0.4.tar.gz: no changes according to diff -ru
• RWeka_0.4-37.tar.gz: changed package description:

  $ diff -ru RWeka.orig RWeka
  diff -ru RWeka.orig/DESCRIPTION RWeka/DESCRIPTION
  --- RWeka.orig/DESCRIPTION	2018-01-07 16:27:10.000000000 +0100
  +++ RWeka/DESCRIPTION	2018-05-23 11:45:56.000000000 +0200
  @@ -26,12 +26,12 @@
   License: GPL-2
   NeedsCompilation: no
   Packaged: 2018-01-07 15:04:47 UTC; hornik
  -Author: Kurt Hornik [aut, cre] (0000-0003-4198-9911),
  +Author: Kurt Hornik [aut, cre] (<https://orcid.org/0000-0003-4198-9911>),
     Christian Buchta [ctb],
     Torsten Hothorn [ctb],
     Alexandros Karatzoglou [ctb],
     David Meyer [ctb],
  -  Achim Zeileis [ctb] (0000-0003-0918-3766)
  +  Achim Zeileis [ctb] (<https://orcid.org/0000-0003-0918-3766>)
   Maintainer: Kurt Hornik <Kurt.Hornik@R-project.org>
   Repository: CRAN
   Date/Publication: 2018-01-07 16:17:29
  diff -ru RWeka.orig/MD5 RWeka/MD5
  --- RWeka.orig/MD5	2018-01-07 16:27:10.000000000 +0100
  +++ RWeka/MD5	2018-05-23 11:45:56.000000000 +0200
  @@ -1,5 +1,5 @@
   5ee28414fe580928907527d9e4217845 *CHANGELOG
  -4aae74779d3a1de0fdc64beec22078ee *DESCRIPTION
  +fe0f10b7f193e91112c978228acaa5ae *DESCRIPTION
   41b1dde3a37014e3c2c5fa208fc47167 *NAMESPACE
   f9a81f720aebf5398a94efa32a2047a5 *R/AAA.R
   e8b6adbe6a0b2cf61f433762e1fd16dd *R/arff.R
  

[boegel]

Test report by @boegel
SUCCESS
Build succeeded for 3 out of 3 (3 easyconfigs in this PR)
node2646.swalot.os - Linux centos linux 7.4.1708, Intel(R) Xeon(R) CPU E5-2660 v3 @ 2.60GHz, Python 2.7.5
See https://gist.github.com/262ff9dae7f4474bb6e6bb1588da8761 for a full test report.

[boegel]

Test report by @boegel
SUCCESS
Build succeeded for 3 out of 3 (3 easyconfigs in this PR)
node2056.delcatty.os - Linux centos linux 7.4.1708, Intel(R) Xeon(R) CPU E5-2670 0 @ 2.60GHz, Python 2.7.5
See https://gist.github.com/c63dc1b2bae40acc87eab7f0bc7148b4 for a full test report.

[boegel]

@JoFAM Thoughts on this, how can this happen?

[JoFAM]

@boegel For pkgmaker version 0.22 was the version on CRAN at the release of R3.5, and due to the drastic internal changes all packages needed to be recompiled repackaged?. According to the archive, this happened on may 10, 2018 :

https://cran.r-project.org/src/contrib/Archive/pkgmaker/

The new version (0.27) is released on may 25, but then the old version was already rebuilt repackaged(?) for the new R version.

I didn't check, but I'm 99% sure that this is the issue for all the packages you mentioned that show no change. This will likely be in the compilation.

edit: I was obviously clearly wrong here

For RWeka it looks like CRAN or Kurt himself corrected the DOI link - they need to have the https:// etc in front. It's something I don't like, but it does happen more often that very small changes in the DESCRIPTION file are made without upping the version.

[boegel]

@JoFAM How does rebuilding for a new R release affect source tarballs? That's just wrong imho...

[JoFAM]

@boegel you're right, and eg pkgmaker doesn't even have compiled code. (shame on me!) Culprit seems to be the tools used. Afaik, CRAN uses gzip to create the tarballs, and gzip adds a timestamp of the creation of the tarball, which results in a different checksum even when the exact same folder is zipped at a different moment.

See eg : https://serverfault.com/questions/110208/different-md5sums-for-same-tar-contents

Looking at the time stamps on the server, it seems like the current version is repackaged on may 10, 2018. I'll check with the powers-that-be if that could epxlain this and if there's any chance of preventing this from happening in the future.

[boegel]

Please do, because this is getting really annoying, and it looks bad imho, checksums for released source tarballs should never change...

[JoFAM]

Done : https://stat.ethz.ch/pipermail/r-package-devel/2018q2/002813.html

I got answer from one of the CRAN maintainers in the meantime. The habit of changing description files in releases is unlikely to change soon. I've asked if there's a possibility to at least avoid checksum changes when archiving. You can follow the discussion using the web interface if you like. Keep in mind there's a little delay on the archive compared to the actual mailing list.

[boegel]

Thanks a lot for raising this @JoFAM!

Even changing the descriptions should result in some kind of version bump imho, for example with an additional digit in the version number like 1.2.3-1 or so...

[JoFAM]

@boegel I wholeheartedly agree, but I'm afraid I don't carry enough weight to make that change happen...