This is the mail archive of the
mailing list for the binutils project.
Re: [RFC PATCH, ARM 5/7] Add support for ARMv8-M security extensions instructions
- From: Thomas Preudhomme <thomas dot preudhomme at foss dot arm dot com>
- To: Nick Clifton <nickc at redhat dot com>
- Cc: Alan Modra <amodra at gmail dot com>, Christophe Lyon <christophe dot lyon at linaro dot org>, binutils at sourceware dot org
- Date: Wed, 11 May 2016 14:32:36 +0100
- Subject: Re: [RFC PATCH, ARM 5/7] Add support for ARMv8-M security extensions instructions
- Authentication-results: sourceware.org; auth=none
- References: <001f01d13873$ff851000$fe8f3000$ at foss dot arm dot com> <20160511051730 dot GB27896 at bubble dot grove dot modra dot org> <ff60ebff-b2df-c03a-424b-13516c8b5b1e at redhat dot com>
On Wednesday 11 May 2016 09:11:22 Nick Clifton wrote:
> Hi Thomas,
> Just noticed this too.
> I have checked in a patch to skip these tests for the old PE/AOUT
> targets for now, but you might want to spend a little time figuring
> out why the tests fail.
Thanks for the fix. The reason for the failure is that objdump only uses
symbols to determine if code is ARM or Thumb whereas it first look at mapping
symbols ($t and $a) for ELF. The testcase does not have any label so the only
symbols are .text, .dada and .bss which is why the test fails on non ELF
target. Fixing that, there's also the fact that objdump prints one nop passed
the last instruction if the alignment is not right. I'm guessing it's due to
how the size is expressed in other formats but I'll go check that.
I'll propose a patch shortly and reenable all these tests.