GNU R related packages: FTBFS in Raspbian and Debian armhf

[mpthompson]

The GNU R related packages are failing on Raspbian with the following issue:

gcc -std=gnu99 -shared -o MASSSHLIB_EXT MASS.o lqs.o SHLIB_LIBADD -L/usr/lib/R/lib -lR
gcc: error: SHLIB_LIBADD: No such file or directory
make[1]: *** [MASSSHLIB_EXT] Error 1
make[1]: Leaving directory `/«PKGBUILDDIR»/src'
make[1]: Entering directory `/«PKGBUILDDIR»/src'
make[1]: Leaving directory `/«PKGBUILDDIR»/src'
ERROR: compilation failed for package 'MASS'
* removing '/«PKGBUILDDIR»/debian/r-cran-mass/usr/lib/R/library/MASS'
make: *** [R_any_arch] Error 1

Full log file can be found here:

http://repo.raspbian.lan/raspbian/logs/main/r/r-cran-mass.log

This problem seems to be inherited from Debian armel and armhf as indicated in this email thread:

http://lists.debian.org/debian-arm/2012/04/msg00015.html
https://buildd.debian.org/status/package.php?p=r-cran-mass&suite=sid

We should track this issue to a resolution on the Debian side of things.

The full list of GNU R packages that fail to compile are:

cairodevice
cluster
date
foreign
gtools
lattice
mnormt
mvtnorm
polspline
quadprog
r-bioc-biobase
r-bioc-limma
r-cran-amore
r-cran-bayesm
r-cran-colorspace
r-cran-digest
r-cran-epi
r-cran-evd
r-cran-gam
r-cran-haplo.stats
r-cran-hdf5
r-cran-int64
r-cran-lpsolve
r-cran-maps
r-cran-mass
r-cran-medadherence
r-cran-multicore
r-cran-nnet
r-cran-plyr
r-cran-randomforest
r-cran-rsqlite
r-cran-slam
r-cran-spatial
r-cran-spc
r-cran-vgam
r-cran-xml
raschsampler
rcpp
rgl
rgtk2
rjava
rmpi
rmysql
robustbase
rodbc
rpvm
rsprng
rsymphony
survival
tkrplot

[Albert Chin (bugs-ubuntu-vendor)]

This bug has been reported to the R-project: https://bugs.r-project.org/bugzilla3/show_bug.cgi?id=15087. Debian has encountered it on arm as well: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=679180.

[peter green (plugwash)]

This has been sorta-fixed in debian

by sorta-fixed I mean that the functionality that was causing the build failure was switched from the internal regex engine to the pcre one. This fixes the build failures but the underlying cause (broken internal regex implementation) is still there.