Hi Jim, Here the a new patch with the following changes according to your suggestions. Jim Blandy wrote: > Why do you need the changes to configure.host, and the bfin.mh file? > You can't run GDB itself on the blackfin processor, can you? It > doesn't have an MMU, and I've never heard of GDB running on uClinux. > The changes to configure.host and bfin.mh have been removed. > In bfin_linux_pc_in_sigtramp, you should pass in the frame and use > safe_frame_unwind_memory instead of deprecated_read_memory_nobpt. It > looks to me like you always have a frame handy --- is that right? > The deprecated_read_memory_nobpt has been replaced with safe_frame_unwind_memory. > + frame_unwind_register (next_frame, BFIN_SP_REGNUM, buf); > + sp = extract_unsigned_integer (buf, 4); > > Is it possible to simply use frame_unwind_register_unsigned in places like this? > Use frame_unwind_register_unsigned in all possible places. > Am I reading bfin_frame_prev_register correctly when I conclude that > the only saved registers it can find are the PC and the FP? How are It can also find other registers besides the PC and the FP. It looks like: if(regnum == BFIN_PC_REGNUM) { ... } else if (regnum == BFIN_FP_REGNUM) { ... } else read_memory (...); > the test suite results? > The summary of the test suite result on simulator is: === gdb Summary === # of expected passes 9113 # of unexpected failures 39 # of expected failures 41 # of known failures 27 # of unresolved testcases 7 # of untested testcases 13 # of unsupported tests 38 The gdb.sum is attached. In the testing, I'm using bfin-elf-gcc (GCC) 4.1.0 20051212 (prerelease) GNU assembler 2.16.91 20051212 GNU ld version 2.16.91 20051212 Newlib comes from . It has not been in the official CVS. > +#include "bfd-in2.h" > + > > Why is this needed? Shouldn't the #include "bfd.h" in defs.h be sufficient? > This include has been removed. > Other than that, it looks nice. > Thanks, Jie