build R package failing with rJava LoadLibrary "not a valid win32 application" failure

I wrote an R package to connect R with messageQueues, specifically activeMQ and rabbitMQ, so all the source can be found here: http://r-forge.r-project.org/projects/r-message-queue/

r-forge fails to build the package, so I'm recreating/debugging locally, but my local build fails earlier in the process than r-forge.

I'm running 64bit windows7, 64bit R v2.15.2.
If I'm going about this the wrong way, let me know.

BUILDING the R package

$ R --verbose CMD build messageQueue
* checking for file 'messageQueue/DESCRIPTION' ... OK
* preparing 'messageQueue':
* checking DESCRIPTION meta-information ... OK
* checking for LF line-endings in source and make files
* checking for empty or unneeded directories
* building 'messageQueue_0.1.tar.gz'
cygwin warning:
  MS-DOS style path detected: C:/Users/msm336/workspace_r/messageQueue_0.1.tar.gz
  Preferred POSIX equivalent is: /cygdrive/c/Users/msm336/workspace_r/messageQueue_0.1.tar.gz
  CYGWIN environment variable option "nodosfilewarning" turns off this warning.
  Consult the user's guide for more details about POSIX paths:
    http://cygwin.com/cygwin-ug-net/using.html#using-pathnames

CHECKING the R package

$ R --verbose CMD check messageQueue_0.1.tar.gz
* using log directory 'C:/Users/msm336/workspace_r/messageQueue.Rcheck'
* using R version 2.15.2 (2012-10-26)
* using platform: x86_64-w64-mingw32 (64-bit)
* using session charset: ISO8859-1
* checking for file 'messageQueue/DESCRIPTION' ... OK
* checking extension type ... Package
* this is package 'messageQueue' version '0.1'
* checking package namespace information ... OK
* checking package dependencies ... OK
* checking if this is a source package ... OK
* checking if there is a namespace ... OK
* checking for executable files ... OK
* checking whether package 'messageQueue' can be installed ... ERROR
Installation failed.
See 'C:/Users/msm336/workspace_r/messageQueue.Rcheck/00install.out' for details.

CONTENTS of messageQueue.Rcheck/00install.out

* installing *source* package 'messageQueue' ...
** R
** inst
** preparing package for lazy loading
** help
*** installing help indices
** building package indices
** testing if installed package can be loaded
*** arch - i386
Error : .onLoad failed in loadNamespace() for 'rJava', details:
  call: inDL(x, as.logical(local), as.logical(now), ...)
  error: unable to load shared object 'C:/Program Files/R/R-2.15.2/library/rJava/libs/i386/rJava.dll':
  LoadLibrary failure:  %1 is not a valid Win32 application.

Error : package 'rJava' could not be loaded
Error: loading failed
Execution halted
*** arch - x64
ERROR: loading failed for 'i386'
* removing 'C:/Users/msm336/workspace_r/messageQueue.Rcheck/messageQueue'

I'm not sure why it's trying to load the i386 version of rJava, when I'm running x64 R on Windows7 x64 using a 64bit version of java:

$ java -version
java version "1.7.0_04"
Java(TM) SE Runtime Environment (build 1.7.0_04-b22)
Java HotSpot(TM) 64-Bit Server VM (build 23.0-b21, mixed mode)

I've also got the JAVA_HOME/jre/bin/server/jvm.dll on my path, because rJava likes that somehow.

Any help would be greatly appreciated.

Thanks,

Matt.

Solution

By default, R's check function will try to test the package against all relevant architectures. On a windows7 x64 machine, this means it will try testing the package against both i386 and x64.

Unfortunately, when running R x64 with a x64 JDK, it seems to not manage architecture dependencies cleanly, thus tries to load a i386 DLL (/rJava/libs/i386/rJava.dll) within an x64 context and fails. I don't fully understand why, but can work around it.

Forcing R CMD check messageQueue to run against the current (default) architecture can be done using the --no-multiarch switch. It is documented in the Writing R Extensions manual, Section 1.3.1 (see: Multiple sub-architectures near the end).

My updated R command to check only against the current architecture:

R --verbose CMD check --no-multiarch messageQueue_0.1.tar.gz

Awesome.