On October 14, 2014 1:29:50 AM CDT, Sebastian Huber <sebastian.hu...@embedded-brains.de> wrote: >On 13/10/14 17:48, Joel Sherrill wrote: >> >> On 10/13/2014 10:43 AM, Sebastian Huber wrote: >>> On 10/13/2014 05:18 PM, Joel Sherrill wrote: >>>> NOTE: For defaults for weak symbols, I am just adding >>>> a prototype to the C file. >>> How do you then ensure that a non-weak implementation has the same >>> signature? >>> >> Don't ask me. I didn't add any of these weak symbol options to any >BSP. >> I am only fixing warnings on existing code. :) > >Adding a prototype for a global function to the C file doesn't fix the >warning. > It just silences the warning and obscures things.
Every case I have seen so far is for bsp_start() to have a weak default version. There is a prototype for that in BSP/bootcard.h and the default version can be static. I have started making the week default static. I will sweep to ensure they all are but if someone beats me to it, I won't complain. >> >> On a more serious note, I would assume that only the linker can >ensure >> this. In ELF, I think it can match the signature. I have no idea if >it >> does or >> not. Create a test case and if it doesn't do check the signature, >post it >> as a question to binutils. >> >> I don't think we have anyway to ensure the match inside our space. >> > >Weak implementations are nothing special. We need a declaration in a >header >file (without the weak attribute). _______________________________________________ devel mailing list devel@rtems.org http://lists.rtems.org/mailman/listinfo/devel
