Re: Has anyone successfully bootstrapped gcc-4.6.3 on m68k?
On 2012-05-18, at 1:54 PM, Vincent Rivière wrote:
> On 18/05/2012 02:05, Vaugha Brewchuk wrote:
>> ../../../../src/gcc-4.6.3-NS-1/libgcc/../gcc/unwind-pe.h:271:1: internal compiler error: in extract_insn, at recog.c:2109
>
> This bug looks similar to that one:
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50882
>
> There are a lot of open bugs in GCC :-(
> IMHO this is not your fault.
>
> In such a case, you can try to manually compile the offending file with the exact same command line (copy/paste from the build log), but by decreasing the optimization level (for example, -O1). Sometimes it works, then you can continue to make normally.
Unfortunately my rebuild with #define DWARF2_UNWIND_INFO 0 died with the same issue while compiling unwind-dw2-fde.c. I tried reducing optimisation options down to no optimisation with no success. I then forced the build process thorugh by ignoring this and the new compiler ICEs while compiling the first libiberty object.
Following Thorsten's suggestions I then applied the following patches from debian:
m68k-fp-cmp-zero.diff
m68k-revert-pr45144.diff
pr43804.diff
pr47612.diff
pr47955.diff
pr48830.diff
pr49169.diff
pr49756.diff
pr52870.diff
svn-updates.diff
Restarted the build about 7 hours ago, so it's "stuck" on genattrtab at the moment. If this build fails, I think I will go back to 4.4.
>> And hence my question: has anyone successfully bootstrapped gcc-4.6.3 on
>> m68k?
>
> I don't bootstrap, I only build a cross-compiler. Less trouble.
> Anyway, the bug happens in libgcc, so the situation is the same since it is built in any case.
>
> I see that your problem comes from unwind-dw2.c. This is about DWARF2 exception unwinding. I didn't hit that bug because I use SJ/LJ exceptions instead (due to lack of ELF support on my favorite platform).
NEXTSTEP does not support ELF either and from reviewing the internals documentation it was not clear to me that I needed to #define DWARF2_UNWIND_INFO 0.
> The right solution would be to post a simplified, preprocessed test case in GCC's Bugzilla. Unfortunately, GCC people does not seem to care much about m68k issues :-(
On the other hand the support here has been incredible! THANK YOU to everyone!
>
> Good luck.
>
> --
> Vincent Rivière
>
>
> --
> To UNSUBSCRIBE, email to debian-68k-REQUEST@lists.debian.org
> with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
> Archive: [🔎] 4FB68CF3.8000606@freesbee.fr">http://lists.debian.org/[🔎] 4FB68CF3.8000606@freesbee.fr
>
Reply to: