Compiling with xcode and --enable-amd64-51-30k has invalid .section in consts.S

[safeandeffectivemyass]

Hi,

I think that there should be some ifdef for compiling with xcode and ./configure --enable-amd64-51-30k, as it gives:

ed25519/amd64-51-30k/consts.S:1:17: error: unexpected token in '.section' directive
.section .rodata

make: *** [ed25519/amd64-51-30k/consts.S.o] Error 1

I fixed it on my end with:

.text

.global .rodata

.globl CRYPTO_NAMESPACE(batch_REDMASK51)
...

Instead of the .section that's there.

Thanks.

[cathugger]

have you tried substituting .section .rodata with .const or .data instead?
it seems that it changed when i ported this to new SUPERCOP, at f374555
used to be .data, i can't really find whether linux supports this but .const seems like it would be MacOS equivalent to what .section .rodata should do on linux.
i guess i could either do conditional compile to use .const for macos, or .data for all targets.

[cathugger]

i don't have a mac to test with

[safeandeffectivemyass]

Let me check.

[safeandeffectivemyass]

That gives me:

ed25519/amd64-51-30k/consts.S:1:8: error: unexpected token in section switching directive
.const .rodata
       ^
make: *** [ed25519/amd64-51-30k/consts.S.o] Error 1

Did I get you right on .const .rodata?

[cathugger]

no, just .const, not .const .rodata

[safeandeffectivemyass]

That works (with just .const).

[cathugger]

could you try replacing it with

#ifndef __APPLE__
.section .rodata
#else
.const
#endif

and tell if that works

[safeandeffectivemyass]

Works well. Thank you.

[safeandeffectivemyass]

Guess this should be applied to amd64-64-24k etc., as well.

[cathugger]

yeah.

[cathugger]

68928c4 should probably be good.

[safeandeffectivemyass]

I'll re-clone and compile, just for good measure.

[safeandeffectivemyass]

Compiled with four different flags on latest and all pass. Thank you!