Re: Bug#1121498: libreoffice: Preprocessor macros in pdfium confuse arm64 with sparc64

Hi Rene,

On Tue, 2025-12-02 at 17:53 +0100, Rene Engelhard wrote:
> Am 27.11.25 um 18:32 schrieb John Paul Adrian Glaubitz:
> > No, the build log is confusing here. It's definitely in pdfium. I have =
applied the
> > patch locally and the build continues past this error now.
> >=20
> > You can see the bug here in the source code [1]. The preprocessor check=
 is
> > incorrect and the reason why the the inline assembly code in [2] is emi=
tted:
> >=20
> > 	asm("qsub %[result], %[first], %[second]"
> >=20
> > As shown above __arch64__ is incorrect for AArch64 since it's supposed =
to be __aarch64__.
>=20
> Apparently that's wrong. From Ubuntu:
>=20
> 14:27 <ricotz> hi, dieser sparc64 fix macht nun aber arm64 kaputt?
>  =C2=A0https://salsa.debian.org/libreoffice-team/libreoffice/libreoffice/=
-/commit/526f26e484aed8c398b44ee6a7dbcd374e89d7f4
> 14:29 <ricotz>=C2=A0 =C2=A0 91 |=C2=A0 =C2=A0 =C2=A0asm("qsub %[result], =
%[first], %[second]"
> 14:29 <ricotz>=C2=A0 =C2=A0 =C2=A0 =C2=A0|=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0^
> 14:29 <ricotz> <inline asm>:1:2: note: instantiated into assembly here
> 14:29 <ricotz>=C2=A0 =C2=A0 =C2=A01 |=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0qs=
ub x22, x9, x8
> 14:29 <ricotz>=C2=A0 =C2=A0 =C2=A0 =C2=A0|=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=
=A0^
> 14:29 <ricotz> workdir/UnpackedTarball/pdfium/core/fxcrt/numerics/safe_ma=
th_arm_impl.h:91:9: error: unrecognized instruction mnemonic, did you mean:=
 fsub,
>  =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0msub, sqsub, sub,=
 subs, uqsub?
> 14:43 <ricotz> (__arch64__) sollte wohl entfernt werden - https://pdfium.=
googlesource.com/pdfium/+/refs/heads/main/core/fxcrt/numerics/safe_conversi=
ons.h
> 17:42 <_rene_> hatte das noch nicht auf arm getestet
> 17:42 <_rene_> schreib das mal in den bug an Adrian
> 17:45 <_rene_> s/bug/bug und/
>=20
> 17:49 <ricotz> https://git.launchpad.net/~libreoffice/ubuntu/+source/libr=
eoffice/commit/?h=3Dubuntu-resolute-25.8&id=3D6099cd68b729afc128ecf8fe3f2da=
d15d7569e32

OK, I'll drop __arch64__ altogether then. I'm still struggling to fix my Ge=
rrit access
in the mean time to fix this issue upstream in pdfium.

I'm surprised though what the original intention of the author of these lin=
es were then
if not to enable ARM-optimized code on AArch64. There is no other 64-bit va=
riant of ARM
that I know of, is there?

Adrian

--=20
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer
`. `'   Physicist
  `-    GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913