Hi Richard,

Many thanks for your explanation.

For now i think is better to drop this pacth and will resend it when I have
a solution.

Richard Purdie <richard.purdie@linuxfoundation.org> escreveu no dia terça,
26/01/2021 à(s) 00:04:

> On Mon, 2021-01-25 at 23:52 +0000, Jose Quaresma wrote:
> > Hi Richard,
> >
> > librt.so is part of the glibc and are allays present on staging
> > sysroot.
> > how does the cmake can find the librt.so outside of the sysroot if it
> > are present on that.
> >
> > The problem is at:
> >
> https://github.com/KhronosGroup/SPIRV-Tools/blob/b27b1afd12d05bf238ac7368bb49de73cd620a8e/source/CMakeLists.txt#L397
> >
> > Probably I need to force the find_library to search first on
> > STAGING_DIR_HOST
>
> I think somehow we need to replace the full sysroot path with a
> variable, or make cmake do that itself if we can. It may be possible to
> have cmake evaluate the variable with reference to the sysroot too but
> that is likely more complex. I'm afraid cmake isn't one of my strong
> areas.
>

I will look at this to see what can be changed in the cmake.bbclass to
achieve
your suggestion.


>
> We do have a second issue too the autobuilder has shown up:
>
>
> https://autobuilder.yocto.io/pub/repro-fail/oe-reproducible-20210125-3iawhiyz/packages/diff-html/
>
> Does something in glslang or spriv-tools use precompiled headers? We
> had problems with that in the vulkan-samples recipe that looked like
> this (a cmake based C++ project where there were changes in the
> debug_loc section) so we may need to disable PCH here too...
>

Yes, I think there are the glslang that have precompiled headers enabled.
I will investigate this issue and send a fix for it.


>
> Cheers,
>
> Richard
>
>

-- 
best regards,
José Quaresma