Some libstdc++-v3 tests in gcc fail because kernel-headers on s390x has in /usr/include/asm/types.h
} __attribute__((packed, aligned(4))) __vector128;
see
https://kojipkgs.fedoraproject.org/packages/gcc/14.0.1/0.15.fc40/data/logs/s390x/build.log
FAIL: 17_intro/headers/c++1998/all_attributes.cc   (test for excess errors)
FAIL: 17_intro/headers/c++2011/all_attributes.cc   (test for excess errors)
FAIL: 17_intro/headers/c++2014/all_attributes.cc   (test for excess errors)
FAIL: 17_intro/headers/c++2017/all_attributes.cc   (test for excess errors)
FAIL: 17_intro/headers/c++2020/all_attributes.cc   (test for excess errors)
These tests attempt to verify that libstdc++ headers themselves never use __attribute__((packed)) etc., but __attribute__((__packed__)) etc., because the former are in the implementation namespace.
grep __attribute__ `rpm -ql kernel-headers` 2>/dev/null | grep -v '((__'
shows hundreds of hits (of course, in some headers it is far more important than in others, the more important ones are those typically included by glibc or libstdc++ headers directly or indirectly).
I think kernel folks don't really want to switch to using those __ prefixed and suffixed versions, could this be done by some sed/awk/python script done when preparing the headers for shipping as user asm/ and linux/ headers?

grep __attribute__ `rpm -ql kernel-headers` 2>/dev/null | grep -v '((__' | sed 's/^.*ttribute__[[:blank:]]*((//;s/))\([^)]*\|$\)//' | sort -u
aligned(256);
aligned(4);
aligned(4) packed_ulong;
aligned(4 * sizeof(__u64));
aligned(8)
aligned(8);
aligned(8) *config;
aligned(8) timestamp_ns;
aligned (__alignof__(struct ebt_replace));
aligned(sizeof(__u64));
aligned(VRING_AVAIL_ALIGN_SIZE)
aligned(VRING_DESC_ALIGN_SIZE)
aligned(VRING_USED_ALIGN_SIZE)
bitwise
packed
packed, aligned(4);
packed,aligned(4)
packed, deprecated

shows that it is really just a couple of attributes and uglifying their names shouldn't be that hard inside of __attribute__(( ... )) arguments.

Reproducible: Always