Hi,
as i had the same issue regarding 1.5.6-r2 not working within eclipse build environment, i felt like having to dig deeper into it.
I used my Segger JLink + free WinIDEA debugging environment from iSystem to see whats going wrong there.
It was pretty obvious whats going wrong there -> it hung in Default_Handler function after entering main and polling a few times for the next SysTick gettin passed.
By comparing the differences in the baselines 1.5.1 vs. 1.5.6 i found the following relevant commit:
which seems to address already a weak symbol definition problem.
Unfortunately this fixture rendered the build in eclipse useless. IMHO it seems like the eclipse build env. does building + linking in another 'order' as it is done by arduino ide.
There are two times weak symbol declarations ( or in case of i.e. the SysTick_Handler even function definitions ) which appear overwriting each others weak alias depending on the link time usage of the objects. So it happens that all declarations (not only the failing SysTick_Hanlder) from cortex_handler.c gettin overwritten by weak symbol redeclaration inside startup_sam3xa.c.
My dirty solution is to just comment out the part in startup_sam3xa.c with the weak symbol declaration regarding the vectors (lines 53 including 128).
It worked also by removing the unchanged startup_sam3xa.c from static library archive build to the project build.
It may also be ok to just remove the files completely (startup handler code + vector map provided by gcc arm built-in libraries is used instead).