Linking (ld) syntax and Floating Point crosscompile question

Caesurus · 06-08-2009, 08:36 AM

Hi,

I'm working on compiling some code from a vendor. They provide a library to link to which I presume was compiled with a gcc 2.9 cross compiler.

I'm having trouble understanding something. Any help or hints would be greatly appreciated. Here is the problem...

The library is supposedly compiled with hard-floating point:

Code:

$ arm-linux-gnu-objdump -p mylib.lib
mylib.lib:     file format elf32-littlearm
private flags = 0: [APCS-32] [FPA float format]

my driver .o files are compiled with soft-float

Code:

$ arm-linux-gnu-objdump -p sm_drv.o
sm_drv.o:     file format elf32-littlearm
private flags = 200: [APCS-32] [FPA float format] [software FP]

Now... when I try to link the two files I get the following, very understandable, error:

Code:

$ arm-linux-gnu-ld -r -o test.o sm_drv.o mylib.lib
arm-linux-gnu-ld: ERROR: mylib.lib uses hardware FP, whereas test.o uses software FP
arm-linux-gnu-ld: failed to merge target specific data of file mylib.lib

BUT... and here is what I don't understand, if I link as follows:

Code:

$ arm-linux-gnu-ld mylib.lib -r -o test.o sm_drv.o

I am able to link the two and the resulting test.o file is:

Code:

$ arm-linux-gnu-objdump -p test.o
test.o:     file format elf32-littlearm
private flags = 200: [APCS-32] [FPA float format] [software FP]

Why is this???

I appreciate the fact that I can ask the vendor to recompile the library for me with my toolchain and my compile flags, but that's not the question here.

Any clues would be greatly appreciated.

Thanks,