[PATCH] elfedit: add support for editing e_flags

[Xi Ruoyao]

On Thu, 2023-03-02 at 19:02 +0800, WANG Xuerui wrote:
> Indeed the object files are correctly LP64S with FP instructions. Won't 
> have to ask every LoongArch distro to upgrade toolchains any more...

You may need to fix and/or upgrade Clang if you want to build the DCN
driver with it.  A Clang "14.0.6" (no idea what it really is, IIRC even
Clang 15 does not support LoongArch) on gcc400.fsffrance.org (Arch Linux
installed) produces:

[xry111@gcc400 ~]$ clang t.c -c -mabi=lp64s -mfpu=64
[xry111@gcc400 ~]$ readelf -a a.out | grep FLOAT
  Flags:                             0x43, DOUBLE-FLOAT, OBJ-v1

And there is some more strange thing:

[xry111@gcc400 ~]$ clang t.c -c -mabi=lp64s -mfpu=32
[xry111@gcc400 ~]$ readelf -a a.out | grep FLOAT
  Flags:                             0x43, DOUBLE-FLOAT, OBJ-v1

Why would "lp64s" and "fpu=32" lead to double float ?!

Cc += Weining, though I'm not sure if this issue still exists in LLVM
mainline.

-- 
Xi Ruoyao <xry111@xry111.site>
School of Aerospace Science and Technology, Xidian University