[Rd] CRAN Check warnings with GCC 8.1

[Peter Langfelder]

Some time ago I (and some other CRAN package maintainers) got an email
from Brian Ripley (copied below) regarding compiler warnings under gcc
8.1. In my case this concerns package PropClust
(https://CRAN.R-project.org/package=PropClust). The check with gcc 8.1
(https://www.stats.ox.ac.uk/pub/bdr/gcc8/PropClust.out) produced a
warning about an uninitialized variable.

I cannot reproduce the warning under gcc 8.1.1 using standard R CMD
check (with or without --as-cran) under R Under development (unstable)
(2018-05-24 r74781); the check finishes "OK" without any warnings or
notes.

Does anyone have any suggestions as to how to reproduce the warning,
or is this something specific to a particular gcc version? My gcc
version 8.1.1 20180502 (Red Hat 8.1.1-1).

Thanks,

Peter



On Sat, May 5, 2018 at 11:53 PM Prof Brian Ripley <ripley using stats.ox.ac.uk> wrote:
>
> This concerns CRAN packages
>
> Amelia C50 Cubist Cyclops DetSel GENLIB IRISSeismic KSgeneral
> MigClim MonetDBLite Numero OpenMx PBSmapping PSPManalysis
> PropClust RArcInfo RandomFields RandomFieldsUtils RcppMsgPack
> RcppParallel RcppRedis RecordLinkage Rmalschains RnavGraph Rvcg
> RxODE SiMRiv antiword bigrquery bsamGP catnet coxme dbarts
> dggridR divest dpglasso earth epanet2toolkit fs gap geojsonsf
> gglasso graphql hashmap haven hier.part imager iptools jiebaR
> kernlab lpridge lvec mlvocab mongolite nandb ore phreeqc polyclip
> qtbase rbamtools rebmix rexpokit rgdal rioja rlas rpf rphast
> rtfbs s2 scio sdnet sdwd sequoia tmlenet tth udpipe validatejsonr
> velox wrassp
>
> GCC 8.1 (the current release) reports more compilation issues than
> previous versions, the more important of which are reported as check
> warnings by R CMD check (in R 3.5.0 and later) and as 'Additional
> issues' on the CRAN check pages.
>
> GCC 8 is already part of the Fedora Linux 28 release and in Debian
> unstable.  Although you may not have access to it yet, the warnings are
> self-contained and usually easy to understand (often that buffers are
> too small).
>
> Please see the reports for your package(s) and correct ASAP.  For
> packages gap and rexpokit these include a .log file with the complete
> compilation log.
>
> --
> Brian D. Ripley,                  ripley using stats.ox.ac.uk
> Emeritus Professor of Applied Statistics, University of Oxford