On Monday 2016-12-05 15:28, Andreas Schwab wrote:
>On Dez 05 2016, Jan Engelhardt <jeng...@inai.de> wrote:
>
>> On Monday 2016-12-05 14:03, Ben Elliston wrote:
>>>On Mon, Dec 05, 2016 at 01:37:21PM +0100, Andreas Schwab wrote:
>>>
>>>> + aarch64:Linux:*:* | aarch64_be:Linux:*:*)
>>>> + eval $set_cc_for_build
>>>> + if echo __ILP32__ | $CC_FOR_BUILD -E - 2>/dev/null \
>>>> + | grep -q __ILP32__
>>>> + then
>>>> + :
>>>> + else
>>>> + UNAME_MACHINE=${UNAME_MACHINE}_ilp32
>>>> + fi
>>>
>>>Please don't use set_cc_for_build. From config.guess:
>>>
>>># CC_FOR_BUILD -- compiler used by this script. Note that the use of a
>>># compiler to aid in system detection is discouraged as it requires
>>># temporary files to be created and, as you can see below, it is a
>>># headache to deal with in a portable fashion.
>>>
>>>Please find another way to get the ilp32 status. I'm not accepting
>>>new uses of set_cc_for_build, because it is an abomination.
>>
>> Furthermore, ILP32 modes are not supposed to be expressed in the platform
>> tuple. x32 configurations, so far, are simply invoked as
>> ./configure --host=x86_64-linux-gnu CFLAGS=-m32 LDFLAGS=-m32
>> because that's what it is.
>
>That doesn't work. Something must distinguish the ABI.
Looks like Debian uses
./configure --build=x86_64-linux-gnux32
as a discriminator. So perhaps if arm needs this too, then it's
something like aarch64-linux-gnuilp32?
_______________________________________________
config-patches mailing list
config-patches@gnu.org
https://lists.gnu.org/mailman/listinfo/config-patches
