chzong wrote: > Just as you said , the .so file was bad . > It is very strange When I using ld to translate the .o to the .so, I should > never place the .o compile by luajit in the front of other .o . Otherwise the > "OS/ABI" field of ELF header > will be changed to "ARM". The correct value is "UNIX - System V". > This is my link command below: the scimarklib.o is compile by luajit. > > /proj/toolchain/armv7a-vfp-neon-wrs-linux/host-cross/usr/toolchain/bin/arm-wrs-linux-gnueabi-ld > > --sysroot=/proj/toolchain/armv7a-vfp-neon-wrs-linux/sysroots/cortexa9-vfp-neon-wrs-linux-gnueabi > -share -fPIC scimarklib.o -L./ -lluajit-5.1 -o scimarklibarm.so Nowadaye, one should never use ld directly. Always use the compiler frontend, i.e. the cross-gcc in your case. This usually ensures that this doesn't happen (by adding a preamble .o). > By the way , We plan to use Cortex A57 arm cpu on our board . > Will luajit support ARM64 architecture in the next two years? Anybody is welcome to contribute an ARM64 port. --Mike