[Rcpp-devel] Issue Installing Rcpp on Solaris

[Tim Jurka]

Hi Dirk,

Coincidentally, my inquiry was prompted by Prof Ripley who was adamant that Rcpp clears R CMD CHECK for solaris-sparc; I personally do not use Solaris.

Thank you for pointing out the test in debugging.cpp, I hadn't noticed that. I'll try out the fix to see if I can successfully install on our university machines and get back to you soon.

Cheers!
Tim

On Oct 17, 2011, at 2:13 PM, Dirk Eddelbuettel wrote:

> 
> Hi Tim,
> 
> On 17 October 2011 at 13:39, Tim Jurka wrote:
> | Hi Rcpp-devel,
> | 
> | When installing Rcpp on Solaris ( R2.13.1 Platform: i386-pc-solaris2.10 (32-bit) ) using install.packages, I get the following error:
> | 
> | g++ -I/usr/local/lib/R/include -I../inst/include/ -I/usr/local/include    -fpic  -g -O2 -c debugging.cpp -o debugging.o
> | debugging.cpp:36:22: execinfo.h: No such file or directory
> | debugging.cpp: In function `SEXPREC* stack_trace(const char*, int)':
> | debugging.cpp:56: error: `backtrace' was not declared in this scope
> | debugging.cpp:57: error: `backtrace_symbols' was not declared in this scope
> | *** Error code 1
> | make: Fatal error: Command failed for target `debugging.o'
> | ERROR: compilation failed for package ‘Rcpp’
> | 
> | Has anybody encountered this problem before, and is there a known fix? Thank you in advance for all your help!
> 
> Thanks for working on Solaris.  
> 
> You may well be the only non-member of R Core who does so --- Prof Ripley
> usually complains to us when Rcpp breaks there, and we usually just shrug our
> shoulders and point out that our hands are tied: no Solaris machine either
> one of us can access, so no (meaningful) way to work on fixes.
> 
> But more recently, Martyn Plummer (CC'ed) provided two excellent patches
> which helped a lot.  With those, we now have Rcpp 0.9.7, and Solaris builds
> on Prof Ripley's test systems, or at least the x86-based one.  Sparc still
> fails as per the results page linked from CRAN's results page for Rcpp.
> 
> Now, for the backtrace and backtrace_symbols symbols were not declared.
> Grep'ing for this I found
> 
> edd at max:~$ grep -rin backtrace svn/rcpp/pkg/Rcpp/src/*
> svn/rcpp/pkg/Rcpp/src/debugging.cpp:56:    stack_depth = backtrace(stack_addrs, max_depth);
> svn/rcpp/pkg/Rcpp/src/debugging.cpp:57:    stack_strings = backtrace_symbols(stack_addrs, stack_depth);
> svn/rcpp/pkg/Rcpp/src/debugging.cpp:67:    free(stack_strings); // malloc()ed by backtrace_symbols
> edd at max:~$ 
> 
> and indeed, in debugging.cpp I see
> 
> #else // ! (defined(WIN32) || defined(__FreeBSD__) || defined(__NetBSD__) || defined(__OpenBSD__) 
> 
> which is lacking a Solaris test -- I may need to add    || defined(__SUNPRO_CC)   
> test as well.
> 
> Could you try locally if that works?  Maybe Google can suggest another
> Solaris variable to test for too?
> 
> Dirk
> 
> -- 
> "Outside of a dog, a book is a man's best friend. Inside of a dog, it is too
> dark to read." -- Groucho Marx