mail being bounced as spam

mail being bounced as spam

[Richard Earnshaw]

[-- Attachment #1: Type: text/plain, Size: 57 bytes --]

Have the spam filters been made more aggressive recently?

[-- Attachment #2: [Postmaster] Email Delivery Failure.eml --]
[-- Type: message/rfc822, Size: 2114 bytes --]

From: Postmaster <postmaster@arm.com>
To: Richard Earnshaw <Richard.Earnshaw@arm.com>
Subject: [Postmaster] Email Delivery Failure
Date: Fri, 13 Apr 2012 17:50:12 +0100
Message-ID: <4373921-1334335812467@service87.mimecast.com>

This is a delivery failure notification message indicating that
an email you addressed to email address :
-- binutils@sourceware.org

could not be delivered. The problem appears to be :
-- Recipient email server rejected the message

Additional information follows :
-- 552 spam score exceeded threshold (#5.6.1)

This condition occurred after 1 attempt(s) to deliver over
a period of 0 hour(s).

If you sent the email to multiple recipients, you will receive one
of these messages for each one which failed delivery,  otherwise
they have been sent.



[-- Attachment #3: Attached Message --]
[-- Type: message/rfc822, Size: 1947 bytes --]

From: Richard Earnshaw <rearnsha@arm.com>
To: Allen Martin <AMartin@nvidia.com>
Cc: 'Meador Inge' <meadori@codesourcery.com>,  "binutils@sourceware.org" <binutils@sourceware.org>
Subject: Re: [PATCH] ld: add switch to disable use of BLX instructions
Date: Fri, 13 Apr 2012 17:49:51 +0100
Message-ID: <4F88592F.8030001@arm.com>

On 13/04/12 16:06, Allen Martin wrote:
>> If this is moving to a single '[no-]use' option, then why do we need
>> ...
>>
>>> @@ -625,6 +628,10 @@ PARSE_AND_LIST_ARGS_CASES='
>>>        use_blx = 1;
>>>        break;
>>>
>>> +    case OPTION_NO_USE_BLX:
>>> +      no_use_blx = 1;
>>> +      break;
>>> +
>>
>> ... 'use_blx' and 'no_use_blx'.  Seems like we just need one of
>> them.
>>
> 
> Because "use_blx" means "force armv5 style interworking" and "no_use_blx" means "force armv4t style interworking".  If you just set "use_blx" to 0 the linker picks for you.  On a linker configured as "--with-arch=armv7-a --with-cpu=cortex-a9 --with-tune=cortex-a9" it *always* picks armv5 style interworking, even if all object files were compiled as "-mcpu=arm7tdmi -march=armv4t -mthumb-interwork"
> 
> -Allen
> 
> nvpublic
> 
> 
> 

As I've explained on a copy of this patch that you posted to the linaro
lists, this is wrong.  You need to identify why the linker thinks it is
OK to use BLX.  It should only do that if it finds an object (or a
library member) in your link set that already indicates that using BLX
is ok.

R.