Bug 924361 – dev-lang/erlang-26.2.1 fails to compile on x86 with gcc-14: erl_bits.c:551:9: error: invalid conversion from type _Float16 without option -msse2]

• Status: CONFIRMED
• Product: Gentoo Linux

[Description Agostino Sarubbo 2024-02-12 10:38:48 UTC]

https://blogs.gentoo.org/ago/2020/07/04/gentoo-tinderbox/

Issue: dev-lang/erlang-26.2.1 fails to compile.
Discovered on: x86 (internal ref: tinderbox_x86)
System: GCC-14-SYSTEM (https://wiki.gentoo.org/wiki/Project:Tinderbox/Common_Issues_Helper#GCC-14)

Info about the issue:
https://wiki.gentoo.org/wiki/Project:Tinderbox/Common_Issues_Helper#CF0014

[Comment 1 Agostino Sarubbo 2024-02-12 10:38:51 UTC]

Created attachment 884795 [details]
build.log

build log and emerge --info

[Comment 2 Matthew Smith 2024-02-25 14:35:42 UTC]

I think that this may be a regression in GCC 14.

If the target doesn't support _Float16, GCC 13 errors on sizeof(_Float16), while GCC 14 returns 2. Making the configure script detect that _Float16 is supported.

Test files:

> $ cat sizeof.c
> #include <stdio.h>
>
> int main() { printf("%lu\n", sizeof(_Float16)); }

> $ cat cast.c
> int main() {
>     _Float16 a = 0;
>     double b = (double) a;
> }

GCC 13 behaviour:

> $ gcc-13 -mno-sse2 sizeof.c 
> sizeof.c: In function ‘main’:
> sizeof.c:3:37: error: ‘_Float16’ is not supported on this target
>     3 | int main() { printf("%lu\n", sizeof(_Float16)); }
>       |                                     ^~~~~~~~
> $ gcc-13 -mno-sse2 cast.c 
> cast.c: In function ‘main’:
> cast.c:2:3: error: ‘_Float16’ is not supported on this target
>     2 |   _Float16 a = 0;
>       |   ^~~~~~~~

GCC 14 behaviour:

> $ gcc-14 -mno-sse2 sizeof.c
> $ ./a.out 
> 2
> $ gcc-14 -mno-sse2 cast.c 
> cast.c: In function ‘main’:
> cast.c:2:3: error: invalid conversion to type ‘_Float16’ without option ‘-msse2’
>     2 |   _Float16 a = 0;
>       |   ^~~~~~~~
> cast.c:3:3: error: invalid conversion from type ‘_Float16’ without option ‘-msse2’
>     3 |   double b = (double) a;
>       |   ^~~~~~