Zlib patch and gold arm-elf target bug report

Doug Kwan (關振德) dougkwan@google.com
Tue Aug 17 06:57:00 GMT 2010


The test should not be removed as it runs fine with arm-*-*eabi for
Android and ARM Linux using EABI.  Configuring gold for arm-*-elf is
not really supported or tested.  Gold is by default not enabled unless
you configure with the --enable-gold option.

-Doug

在 2010年8月16日下午11:40,Rus <harbour@sfinx.od.ua> 寫道:
> On Mon, 16 Aug 2010, Doug Kwan (關振德) wrote:
>
> :If you want to have a working linker now, I would suggest you using
> :ld, which has better support for the old ABI instead.  When code was
> :ported from BFD, to gold, support for the older ABI was removed to
> :simplify things.  gold can be changed to support the older ELF ABI but
> :it will not be fixed in the very near future.  gold being open-source,
> :you are more than welcome to modify it and submit your patches
> :provided that the maintainer accepts it.
>
>  I do not use gold at all because it is not stable even for x86 arch - the
> ld still remains the only working linker in binutils. But if gold can't
> support the same ABI range as ld does - such regression tests for
> gold need to be removed as standart procedure 'make check' will fail for
> all the binutils package when compiling for such arches.
>
> :-Doug
> :
> :在 2010年8月16日下午10:02,Rus <harbour@sfinx.od.ua> 寫道:
> :> On Mon, 16 Aug 2010, Doug Kwan (關振德) wrote:
> :>
> :>        Hello,
> :>
> :> :Is there any reason why you are using ELF instead of EABI?  The
> :> :arm-elf tool-chain uses some deprecated relocation types, which are
> :> :not handled in gold.  In particular, the deprecated R_ARM_PC24
> :> :relocated type is not handled properly and caused the assert failure.
> :>
> :> My target is embedded ARM7 CPU with 32 Kbytes of FLASH - seems like it is
> :> ArmV3 arch which do not support EABI. Anyway until the binutils claims
> :> that supports arches older than armv4 and arm-elf systems and ld passes
> :> the regression tests - this will remains the gold bug.
> :>
> :> :
> :> :-Doug
> :> :
> :> :在 2010年8月16日上午8:48,Rus <harbour@sfinx.od.ua> 寫道:
> :> :>
> :> :>        Hello,
> :> :>
> :> :> The following patch needed for compiling binutils with Werror :
> :> :>
> :> :> --- binutils-2.20.51.0.11/bfd/compress.c        2010-08-12
> :> :> 00:52:42.000000000 +0300
> :> :> +++ binutils-2.20.51.0.11-1/bfd/compress.c      2010-08-16
> :> :> 18:20:45.004000011 +0300
> :> :> @@ -94,7 +94,7 @@
> :> :>   bfd_set_error (bfd_error_invalid_operation);
> :> :>   return FALSE;
> :> :>  #else
> :> :> -  bfd_size_type compressed_size;
> :> :> +  uLongf compressed_size;
> :> :>   bfd_byte *compressed_buffer;
> :> :>
> :> :>   compressed_size = compressBound (uncompressed_size) + 12;
> :> :>
> :> :>
> :> :> I use 3 arches (x86, avr and arm-ef). 2 first are good, but 'make check' for
> :> :> arm-elf target is failed with :
> :> :>
> :> :> make[5]: `arm_abs_global.stdout' is up to date.
> :> :> ../ld-new -T ./arm_branch_range.t -o arm_bl_in_range arm_bl_in_range.o
> :> :> ../ld-new: internal error in relocate, at arm.cc:8944
> :> :> make[5]: *** [arm_bl_in_range] Error 1
> :> :> make[5]: Leaving directory
> :> :> `/Arhiv/Soft/src/binutils/1/binutils-2.20.51.0.11/gold/testsuite'
> :> :>
> :> :>
> :> :>                Rus
> :> :>
> :> :
> :>
> :>                Rus
> :
>
>                Rus



More information about the Binutils mailing list