On (06/09/12 12:36), Brandon Stafford wrote: > Hi all, > > I'm working on a recipe for Scipy, which is a Python library for > scientific computation. It contains chunks of code in C and Fortran. > > I have seen this recent message from Phil Blundell (about an entirely > different recipe) that suggests that the problem may be the use of the > host's cc and ld, rather than the correct ARM cross-compilers: > http://www.mail-archive.com/[email protected]/msg27120.html > > This looks like the problem to me, but I'm not sure how to tell the > Python build script that it should be cross-compiling. Based on Phil's > message, I'd expect it involves setting ${CC} and ${LD} and somehow > passing them to Python, but I'm not sure how. > > Full error message: http://pastebin.com/pMvaLpcz > Recipe in broken state: http://pastebin.com/a5jm45vB > > (The line export LDFLAGS = "-L${STAGING_LIBDIR}" is just there because > another mailing list suggestion: > http://www.mail-archive.com/[email protected]/msg24196.html. > It does not appear to affect the errors produced.) > > Any suggestions?
FIrst you need fortran language frontends and support libraries compiled with gcc cross compiler. By default we only build c/c++ the you need to take a look at scipy build system and see how its deriving its CC and LD variables and then override them with CC LD that recipe gets from environment when building > > Brandon > > -- > Brandon Stafford > Rascal Micro: small computers for art and science > Somerville, MA, USA > > _______________________________________________ > Openembedded-core mailing list > [email protected] > http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core -- -Khem _______________________________________________ Openembedded-core mailing list [email protected] http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core
