Taeke created ARROW-7908: ---------------------------- Summary: Can't install R-library arrow without setting LIBARROW_DOWNLOAD=true Key: ARROW-7908 URL: https://issues.apache.org/jira/browse/ARROW-7908 Project: Apache Arrow Issue Type: Bug Components: R Affects Versions: 0.16.0 Environment: Operating System: Red Hat Enterprise Linux Server 7.6 (Maipo) CPE OS Name: cpe:/o:redhat:enterprise_linux:7.6:GA:server Kernel: Linux 3.10.0-957.35.2.el7.x86_64 Architecture: x86-64 Reporter: Taeke Fix For: 0.16.0
Hi, Installing arrow in R does not work intuitively on our server. {code:r} install.packages("arrow")` {code} results in an error: {code:sh} Installing package into '/home/<user>/R/x86_64-redhat-linux-gnu-library/3.6' (as 'lib' is unspecified) trying URL 'https://cloud.r-project.org/src/contrib/arrow_0.16.0.2.tar.gz' Content type 'application/x-gzip' length 216119 bytes (211 KB) ================================================== downloaded 211 KB * installing *source* package 'arrow' ... ** package 'arrow' successfully unpacked and MD5 sums checked ** using staged installation PKG_CFLAGS=-I/tmp/Rtmp3v1BDf/R.INSTALL4a5d5d9f8bc8/arrow/libarrow/arrow-0.16.0.2/include -DARROW_R_WITH_ARROW PKG_LIBS=-L/tmp/Rtmp3v1BDf/R.INSTALL4a5d5d9f8bc8/arrow/libarrow/arrow-0.16.0.2/lib -larrow_dataset -lparquet -larrow -lthrift -lsnappy -lz -lzstd -llz4 -lbrotlidec-static -lbrotlienc-static -lbrotlicommon-static -lboost_filesystem -lboost_regex -lboost_system -ljemalloc_pic ** libs g++ -m64 -std=gnu++11 -I"/usr/include/R" -DNDEBUG -I/tmp/Rtmp3v1BDf/R.INSTALL4a5d5d9f8bc8/arrow/libarrow/arrow-0.16.0.2/include -DARROW_R_WITH_ARROW -I"/usr/lib64/R/library/Rcpp/include" -I/usr/local/include -fpic -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector-strong --param=ssp-buffer-size=4 -grecord-gcc-switches -m64 -mtune=generic -c array.cpp -o array.o In file included from array.cpp:18:0: ./arrow_types.h:201:31: fatal error: arrow/dataset/api.h: No such file or directory {code} It appears that the C++ code is not built. With arrow 0.16.0.1 things do work out, because it tries to build the C++ code from source. With arrow 0.16.0.2 such is no longer the case. I could finish the installation by setting the environment variable LIBARROW_DOWNLOAD to 'true': {code:java} export LIBARROW_DOWNLOAD=true {code} That, apparently, triggers the build from source. I would have expected that I would not need to set this variable explicitly. I found that [between versions|[https://github.com/apache/arrow/commit/660d0e7cbaa1cfb51498299d445636fdd6a58420]], the default value of LIBARROW_DOWNLOAD has changed: {code:sh} - download_ok <- locally_installing && !env_is("LIBARROW_DOWNLOAD", "false") + download_ok <- env_is("LIBARROW_DOWNLOAD", "true") {code} In our environment, that variable was _not_ set, resulting (accidentally?) in download_ok being false and therefore the libraries not being installed and finally the resulting error above. I can't quite figure out the logic behind all this, but it would be nice if we'd be able to install the package without first having to set LIBARROW_DOWNLOAD. Thank you for looking into this! -- This message was sent by Atlassian Jira (v8.3.4#803005)