Re: Has anyone successfully bootstrapped gcc-4.6.3 on m68k?
(note that I changed my email account from Vaugha to t-rexky to avoid using yahoo)
On 2012-05-20, at 8:57 AM, Thorsten Glaser wrote:
> Vaugha Brewchuk dixit:
>
>> Just a quick update: the same issue persists with the above patches
>> applied. I decided to spend some time with gcc-4.4 for now, and will
>> go back to 4.6 a bit later.
>
> Hm ok, good luck.
Thank you very much to Thorsten and to Finn for links to the Debian patch sources. I attempted to build 4.4.7 but it ICE'd at the same place. I then had another good look at Andreas' post (thank you very much once again Andreas) and the following statement made me experience an eureka moment:
> No, it's a tablejump with CASE_VECTOR_PC_RELATIVE but CASE_VECTOR_MODE
> defined to SImode, which is not a supported configuration.
I looked at my host configuration files and sure enough I think I found the source of the problem. The original NeXT gcc-2.5 configuration files that I used as reference had the following definition in nextstep.h:
#define JUMP_TABLES_IN_TEXT_SECTION 1
and m68k/next.h had the following definitions:
#undef CASE_VECTOR_MODE
#define CASE_VECTOR_MODE SImode
#undef CASE_VECTOR_PC_RELATIVE
There is also a contradictory comment in m68k/next.h that talks about not defining JUMP_TABLES_IN_TEXT_SECTION. Interestingly, the GNU config files from gcc-3.2.3 do not define JUMP_TABLES_IN_TEXT_SECTION. I removed the definition of JUMP_TABLES_IN_TEXT_SECTION and am currently rebuilding 4.4.7. If this succeeds, I will move back to 4.6.3.
I will post another update some time tomorrow.
Thank you to all!
Reply to: