On 23/7/21 5:00 pm, Sebastian Huber wrote: > Hello Chris, > > On 22/07/2021 10:44, Chris Johns wrote: >> Hello, >> >> Libbsd uses the pre-processor to map all the kernel calls into a libbsd >> kernel >> name space by prepending _bsd_ to each symbol. The script ... >> >> https://git.rtems.org/rtems-libbsd/tree/create-kernel-namespace.sh?h=6-freebsd-12 >> >> ... generates the list and the result is pushed into the repo. The symbols >> need >> to be regenerated when new sources are added into the `freebsd` tree. >> >> The script has a few issues: >> >> 1. Objdump does not work on FreeBSD for different archs. >> >> 2. Binutils is being removed from FreeBSD base. > > what would be the alternative?
The ones we build. It is not a big change to the script but one that needs to be amde. We cannot expect the host OS to provide a suitable objdump. >> 3. A number of BSPs need to be built to cover all the possible symbols >> >> I would like to document the list of BSPs a generate needs to cover. I >> propose: >> >> arm/xilinx_zynq_a9_qemu >> aarch64/xilinx_versal_lp64_qemu >> i386/pc686 >> powerpc/mvme2307 >> sparc/erc32 > > Basically if you import code you just have to build a BSP which covers the > imported code. Then you use ... There are other complexities that only a handful of people in the world would know or understand. For example you build a zynq BSP or any 32bit BSP and import code that conditionally defines 64bit functionality then you may miss some symbols. We have cases of this, well I am seeing them adding VFS. >> Also the documentation says to use `git add -p` to add the changes. How does >> an >> interactive add help? > > git add -p > > to add only the changes relevant to the imported (or removed or changed) code. OK and this is good advice so I will add something to the doco to state this is why. A baseline list of BSPs is also needed to regenerate the file. We cannot add for ever because some symbols disappear over time. Chris _______________________________________________ devel mailing list devel@rtems.org http://lists.rtems.org/mailman/listinfo/devel