This is the mail archive of the
gcc-help@gcc.gnu.org
mailing list for the GCC project.
Re: v8 LTO build fails on arm-linux-gnueabihf
- From: Prathamesh Kulkarni <prathamesh dot kulkarni at linaro dot org>
- To: Kyrill Tkachov <kyrylo dot tkachov at arm dot com>
- Cc: Martin LiÅka <mliska at suse dot cz>, gcc-help <gcc-help at gcc dot gnu dot org>
- Date: Thu, 16 Apr 2015 03:01:30 +0530
- Subject: Re: v8 LTO build fails on arm-linux-gnueabihf
- Authentication-results: sourceware.org; auth=none
- References: <CAAgBjMnS4V_4H_p2HQ2194niuoKmwnNjd33sFM+jZLwKf3pZzw at mail dot gmail dot com> <55239546 dot 70109 at arm dot com>
On 7 April 2015 at 13:58, Kyrill Tkachov <kyrylo.tkachov@arm.com> wrote:
>
> On 05/04/15 21:33, Prathamesh Kulkarni wrote:
>>
>> ../test/unittests/compiler/simplified-operator-unittest.cc:23:61: note: a
>> field of same name but different type is defined in another translation unit
>> const Operator* (SimplifiedOperatorBuilder::*constructor)();
>> ^
>> /tmp/cc8drsSQ.s: Assembler messages:
>> /tmp/cc8drsSQ.s:69493: Error: offset out of range
>> /tmp/cc8drsSQ.s:69639: Error: offset out of range
>
>
> Hi Prathamesh,
>
> Last time I saw this error, it was because some define_insn in the arm
> backend had the wrong 'length' attribute. In particular, if the
> 'length' attribute underestimates the true length of the instruction
> then the compiler may try to pack too many instructions into between
> a branch and a label, causing this error. If there is a way to get a
> preprocessed file for this, compiling with -dp to annotate the assembly
> with the define_insn patterns used and the 'length' attribute should
> help with tracking this down, if that is indeed what's the problem.
>
Hi Kyrill,
Sorry for late response, the assembly corresponds to inline assembly
in src/arm/cpu-arm.cc.
I have filed this as PR65778.
Thank you,
Prathamesh
> Kyrill
>