Bug#1054180: openssl: FTBFS on loong64

2023-10-20 Thread WANG Xuerui

On 10/21/23 01:58, Sebastian Andrzej Siewior wrote:

On 2023-10-20 13:49:01 [+0800], WANG Xuerui wrote:

This is likely upstream issue
https://github.com/openssl/openssl/issues/21340, manifesting on Debian due
to the packaged binutils not having Loongson SIMD support yet, and fixed by
upstream commit 160f48941d143cf2682df4e938ba953c96ac3c7a.

I'd suggest including the upstream patch so we don't have to touch anywhere
else.

So submitting patches for projects before having the upstream in the
building tools. Nobody can build this then. What about uploading a
custom binutils with the needed support to ports until it is merged
upstream?


Hmm, not quite; with the commit I linked to applied, the build will get 
fixed, only that the asm code won't get built. It's a matter of the 
original change adding SIMD optimization not taking care of toolchain 
probing.


Also the binutils instruction set support is actually complete and 
released in 2.41, so a custom  build isn't necessary either: we'd only 
have to wait until Debian bumps the binutils version to 2.41 across the 
board. It's a good thing to not move so quickly, on the other hand; 
we've already hit some interesting bugs in Gentoo land due to the linker 
relaxation thing...




Bug#1054180: openssl: FTBFS on loong64

2023-10-20 Thread Sebastian Andrzej Siewior
On 2023-10-20 13:49:01 [+0800], WANG Xuerui wrote:
> Hi,
Hi,

> This is likely upstream issue
> https://github.com/openssl/openssl/issues/21340, manifesting on Debian due
> to the packaged binutils not having Loongson SIMD support yet, and fixed by
> upstream commit 160f48941d143cf2682df4e938ba953c96ac3c7a.
> 
> I'd suggest including the upstream patch so we don't have to touch anywhere
> else.

So submitting patches for projects before having the upstream in the
building tools. Nobody can build this then. What about uploading a
custom binutils with the needed support to ports until it is merged
upstream?

Sebastian



Bug#1054180: openssl: FTBFS on loong64

2023-10-19 Thread WANG Xuerui

Hi,

This is likely upstream issue 
https://github.com/openssl/openssl/issues/21340, manifesting on Debian 
due to the packaged binutils not having Loongson SIMD support yet, and 
fixed by upstream commit 160f48941d143cf2682df4e938ba953c96ac3c7a.


I'd suggest including the upstream patch so we don't have to touch 
anywhere else.




Bug#1054180: openssl: FTBFS on loong64

2023-10-18 Thread Sebastian Andrzej Siewior
Package: openssl
Version: 3.1.2-1
Severity: wishlist
Tags: ftbfs
User:debian-de...@lists.debian.org
Usertags: loongarch64

OpenSSL 3.1.x in experimental FTBFS on loong64:

|crypto/aes/vpaes-loongarch64.S: Assembler messages:
|crypto/aes/vpaes-loongarch64.S:24: Error: no match insn: vori.b
$vr1,$vr9,0
|crypto/aes/vpaes-loongarch64.S:26: Error: no match insn: vld   $vr2,$r12,0
|crypto/aes/vpaes-loongarch64.S:27: Error: no match insn: vandn.v   
$vr1,$vr1,$vr0
…

Latest build log

https://buildd.debian.org/status/fetch.php?pkg=openssl&arch=loong64&ver=3.1.3-1&stamp=1695145526&raw=0

My guess is that either binutils support is missing or a flag/ CPU
switch is missing to recognise the vector opcodes.

Sebastian