Skip to content

[Rcpp-devel] More on macOS and Rcpp 1.0.4

5 messages · Davis Vaughan, Dirk Eddelbuettel, Kasper Daniel Hansen

Original
Davis Vaughan
# 
Follow up of
http://lists.r-forge.r-project.org/pipermail/rcpp-devel/2020-March/010411.html
.

I am on a macOS running Mojave, 10.14.5.

I ran a partial set of revdeps for Rcpp (in alphabetical order), and 39 out
of 546 fail with CRAN Rcpp on macOS because of the following issue, fixed
in Rcpp devel by Kevin: https://github.com/RcppCore/Rcpp/issues/1046

Here is the list of failures from my partial test. Notably (at least to
me), are dplyr and bigmemory.

* AlphaSimR
* BiocNeighbors
* COHCAP
* CancerInSilico
* CoGAPS
* Cyclops
* DCEM
* DNAshapeR
* DNAtools
* DescTools
* DropletUtils
* EDDA
* bamsignals
* biganalytics
* bigmemory
* bigrquery
* bio3d
* bnclassify
* bzinb
* castor
* clifford
* collUtils
* colourvalues
* contiBAIT
* cppRouting
* csaw
* ctgt
* dada2
* datastructures
* detrendr
* dfcomb
* dipsaus
* divest
* dodgr
* dplyr
* dqrng
* dtwclust
* exceedProb
* fastlogranktest

-Davis
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/rcpp-devel/attachments/20200318/8c0960c3/attachment.html>
1 day later
Dirk Eddelbuettel
#
On 18 March 2020 at 10:48, Davis Vaughan wrote:
| Here is the list of failures from my partial test. Notably (at least to
| me), are dplyr and bigmemory.

Two days later. Neither one is failing at CRAN:

  https://cloud.r-project.org/web/checks/check_results_bigmemory.html
  https://cloud.r-project.org/web/checks/check_results_dplyr.html

May still put up a bug-fix release 1.0.4-1 but I am really not too thrilled
by people on social media claiming "widespread failure". I don't see it.

And to be clear, the R 3.3.* issue some of you have ... is very unfortunate
but also outside of our purview. We never claimed to test that far back. We
invited the community to test beyond what we do -- CRAN level plus reverse
depends -- but life is short and all that nobody got to it. It happens. I
simply would not test on R releases older than what CRAN tests but that may
just be me.  I would be happy to work with a volunteer test team that wants
to go beyond what CRAN mandates. Happy to support that, cannot do it myself.

Dirk
1 day later
Dirk Eddelbuettel
# 
As CRAN told me this morning, BioConductor upgraded the GoSemSim package. So
CRAN no longer exeriences any breakage related to the 1.0.4 release, and sees
no need for an immediate patch release.  Neither do I.

For those bitten by whatever failings that revealed itself only outside the
test matrix --- like using R 3.3.* --- we do have Rcpp 1.0.4.3 with its
incremental fixed on the drat as previously stated.  As the question of "why
can't I compile the brandnew Rcpp on my almost four-year old R release" came
up at StackOverflow, it put a few lines together in response which are here:

  https://stackoverflow.com/questions/60783528/compilation-failed-when-installing-rcpp/60783584#60783584

I also mentioned the post on Twitter

  https://twitter.com/eddelbuettel/status/1241180708229582850

and e.g. Kyle Baron followed up with a quoting tweet

  https://twitter.com/kylebtwin/status/1241366335705231361

supporting the main thesis: more contributed testing would be helpful.
Maybe we all can try that for 1.0.5 in a few months.

If other issues come up, please do raise them here or at the GitHub repo.

Cheers, Dirk
Dirk Eddelbuettel
# 
We are now at Rcpp 1.0.4.4 on master and in the drat -- Mattias Ellert
contributed _another_ cleanup PR on the recent exceptions refactoring PR
fixing two overlooked aspects that came up at ROOT.

As before, master is current and the corresponding tar.gz is at the drat:

   install.packages("Rcpp", repos="https://rcppcore.github.io/drat")

With special thanks to Mattias,  Dirk
9 days later
Kasper Daniel Hansen
# 
FYI, based on my experiences today:
  - I am installing the R-4-0 branch from svn today (march 31)
  - I am using clang8 on OS X as recommended by CRAN/Simon
  - I am using homebrew, a widely used package manager on OS X, for many of
my dependencies

1. Before installing R-4-0 I updated homebrew to get the latest versions of
everything as of today.
2. I then compiled R and Rcpp, no errors.
3. However, when compiling RSQLite which depends on BH and Rcpp I get the
error listed below
4. This was fixed by installing Rcpp 1.0.4.4 from the drat archive in the
previous message by Dirk.

So for me - with the very latest homebrew + clang8 on OS X - I found the
fix necessary to install RSQLite from source.

All of this may be noise, I have not had time to do a thorough
investigation. Also, for the non-OS X people out there, CRAN does not use
homebrew to pull in its dependencies, so there's that.

Best,
Kasper


--- Error from installing RSQLIte ---

In file included from DbColumn.cpp:1:
In file included from ./pch.h:1:
In file included from ./RSQLite.h:14:
In file included from
/usr/local/R/4.0/lib/R/site-library/BH/include/boost/container/stable_vector.hpp:23:
In file included from
/usr/local/R/4.0/lib/R/site-library/BH/include/boost/config.hpp:57:
In file included from
/usr/local/R/4.0/lib/R/site-library/BH/include/boost/config/platform/macos.hpp:28:
In file included from
/usr/local/R/4.0/lib/R/site-library/BH/include/boost/config/detail/posix_features.hpp:18:
/usr/include/unistd.h:665:27: error: unknown type name 'uuid_t'; did you
mean 'uid_t'?
int      getsgroups_np(int *, uuid_t);
On Sun, Mar 22, 2020 at 8:58 AM Dirk Eddelbuettel <edd at debian.org> wrote: