RE: [PATCH 1/8] ld: Rename aarch64pe emulation target to arm64pe

["Martin Storsjö" <martin@martin.st>]

[-- Attachment #1: Type: text/plain, Size: 2074 bytes --]

On Wed, 4 Jan 2023, Tamar Christina wrote:

>> -----Original Message-----
>> From: Martin Storsjö <martin@martin.st>
>> Sent: Wednesday, January 4, 2023 10:25 AM
>> To: nickc@redhat.com
>> Cc: Mark Harmstone <mark@harmstone.com>; Tamar Christina
>> <Tamar.Christina@arm.com>; Andrew Pinski <pinskia@gmail.com>; Richard
>> Earnshaw <Richard.Earnshaw@foss.arm.com>; NightStrike
>> <nightstrike@gmail.com>; wej22007@outlook.com; zac.walker@linaro.org;
>> binutils <binutils@sourceware.org>
>> Subject: Re: [PATCH 1/8] ld: Rename aarch64pe emulation target to arm64pe
>>
>> On Wed, 4 Jan 2023, Nick Clifton wrote:
>>
>>> This would create duplicated code sure, but not too much, and it would
>>> allow the linker to be compatible with Clang whilst still also
>>> retaining the aarch64 moniker preferred by ARM.
>>
>> As this still is framed as "compatible with Clang" - does this mean that you still
>> insist that GCC should use a different emulation name when calling the
>> linker, than what Clang does, forcing lld to also start recognizing that new,
>> different emulation name - different from the one that has been in place for
>> 5 years?
>
> To be clear, GCC will very likely reject any port upstreaming that uses 
> Arm64 for the same reason.  And GCC is a lot more tightly controlled. 
> At least if you want upstream support.
>
> Believe it or not I'm actually trying to help you here, the fact that 
> clang has called it arm64 is why we want to allow the alias.  As I 
> mentioned before, it just can't be the main name.

Right - but in the case of GCC, it wouldn't be the name of a new port - it 
would only be a single word in a file for passing linker arguments. Just 
like how "-m i386pep" is passed for x86_64 today here: 
https://github.com/gcc-mirror/gcc/blob/345dffd0d4ebff7e705dfff1a8a72017a167120a/gcc/config/i386/mingw-w64.h#L74

So regardless of the binutils discussion, are you saying that GCC would 
reject a patch that adds such an occurance that would pass "-m arm64pe", 
in a patch that otherwise consistently calls the port "aarch64"?

// Martin