sh architecture, "hello" sample doesn't work

Steve Battazzo sbattazzo at
Fri Jun 25 15:36:39 UTC 2010

Ralf and Alberico,
Thanks for the replies!
> As Ralf said: nearly surely you need to compile with the rtems toolchain and 
> the  memory layout (the linkcmds file )should reflect your hardware memory map.
The rtems application of course is compiled with the rtems toolchain, 
but as the BSP's readme said, the gdbstubs should be compiled with 
sh-elf-gcc, right?

I have been using the precompiled sh-elf-gcc from KPIT GNU tools because 
it has worked so far for me (for bare sh2 code anyway), and I always 
have issues when trying to compile the GCC for this target from source.

The gdbstubs on sourceforge wouldn't compile because of some syntax 
regarding some pointers, and I didn't really want to bother with trying 
to fix it, so I tried the one provided by KPIT, which at least 
successfully connected and loaded the program.
> Stated this, here's the main problems I recall to run "hello.exe" :
> Overlap  of the gdbstub working memory by  loading the program.
In my case, a look at the gdbstub linker script shows no reference to 
the external memory at all, but now that you mention it, there could be 
an issue with the chip's internal memory!
Running objcopy and taking a look at the generated srec file shows some 
data is written to the on-chip RAM region, which might  overwrite some 
of the gdbstub's memory as you suggest.
> Disassemble the start address and see if the program is correctly loaded.
> depending on your gdbstub implementation, you may need to setup some cpu 
> registers to fit the memory configuration before to load the executable.
I will try this next time I get some time to do so.. thank you!



More information about the users mailing list