issue with latest release of R-devel

I'm getting ready to submit an update of survival, and is my habit I run the checks on all
packages that depend/import/suggest  survival.  I am getting some very odd behaviour wrt
non-reproducability.  It came to a head when some things failed on one machine and worked
on another.   I found that the difference was that the failure was using the 3/27 release
and the success was still on a late Jan release.   When I updated R on the latter machine
it now fails too.

An example is the test cases in genfrail.Rd, in the frailtySurv package.   (The package
depends on survival, but I'm fairly sure that this function does not.)   It's a fairly
simple function to generate test data sets, with a half dozen calls in the test file.  If
you cut and paste the whole batch into an R session, the last one of them fails.  But if
you run that call by itself it works.   This yes/no behavior is reproducable.

Another puzzler was the ranger package.  In the tests/testthat directory,
source('test_maxstat') fails if it is preceeded by source('test_jackknife'), but not
otherwise.  Again, I don't think the survival package is implicated in either of these tests.

Another package that succeeded under the older r-devel and now fails is arsenal, but I
haven't looked deeply at that.

Any insight would be be appreciated.

Terry T.
----


Here is the sessionInfo() for one of the machines.  The other is running xubuntu 18 LTS.
(It's at the office, and I can send that tomorrow when I get in.)

R Under development (unstable) (2019-03-28 r76277)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 16.04.6 LTS

Matrix products: default
BLAS:   /usr/local/src/R-devel/lib/libRblas.so
LAPACK: /usr/local/src/R-devel/lib/libRlapack.so

locale:
  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C
  [3] LC_TIME=en_US.UTF-8        LC_COLLATE=C
  [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8
  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C
  [9] LC_ADDRESS=C               LC_TELEPHONE=C
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods base

loaded via a namespace (and not attached):
[1] compiler_3.6.0


        [[alternative HTML version deleted]]

Could this be related to

"SIGNIFICANT USER-VISIBLE CHANGES

The default method for generating from a discrete uniform distribution
(used in sample(), for instance) has been changed. This addresses the
fact, pointed out by Ottoboni and Stark, that the previous method made
sample() noticeably non-uniform on large populations. See PR#17494 for
a discussion. The previous method can be requested using RNGkind() or
RNGversion() if necessary for reproduction of old results. Thanks to
Duncan Murdoch for contributing the patch and Gabe Becker for further
assistance."

If so, testing with

    export _R_RNG_VERSION_=3.5.0

might remove/explain those errors.

Just a thought

Henrik

On Wed, Mar 27, 2019 at 8:16 PM Therneau, Terry M., Ph.D. via R-devel
<r-devel at r-project.org> wrote:

I'm getting ready to submit an update of survival, and is my habit I run the checks on all
packages that depend/import/suggest  survival.  I am getting some very odd behaviour wrt
non-reproducability.  It came to a head when some things failed on one machine and worked
on another.   I found that the difference was that the failure was using the 3/27 release
and the success was still on a late Jan release.   When I updated R on the latter machine
it now fails too.

An example is the test cases in genfrail.Rd, in the frailtySurv package.   (The package
depends on survival, but I'm fairly sure that this function does not.)   It's a fairly
simple function to generate test data sets, with a half dozen calls in the test file.  If
you cut and paste the whole batch into an R session, the last one of them fails.  But if
you run that call by itself it works.   This yes/no behavior is reproducable.

Another puzzler was the ranger package.  In the tests/testthat directory,
source('test_maxstat') fails if it is preceeded by source('test_jackknife'), but not
otherwise.  Again, I don't think the survival package is implicated in either of these tests.

Another package that succeeded under the older r-devel and now fails is arsenal, but I
haven't looked deeply at that.

Any insight would be be appreciated.

Terry T.
----


Here is the sessionInfo() for one of the machines.  The other is running xubuntu 18 LTS.
(It's at the office, and I can send that tomorrow when I get in.)

R Under development (unstable) (2019-03-28 r76277)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 16.04.6 LTS

Matrix products: default
BLAS:   /usr/local/src/R-devel/lib/libRblas.so
LAPACK: /usr/local/src/R-devel/lib/libRlapack.so

locale:
   [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C
   [3] LC_TIME=en_US.UTF-8        LC_COLLATE=C
   [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8
   [7] LC_PAPER=en_US.UTF-8       LC_NAME=C
   [9] LC_ADDRESS=C               LC_TELEPHONE=C
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods base

loaded via a namespace (and not attached):
[1] compiler_3.6.0


         [[alternative HTML version deleted]]