On 1/23/20 8:07 AM, Pétur Orri Ragnarsson wrote:
> Hi.
>
> I ran into a problem with fixincludes when building GCC. Your email is
> in the README file.
>
>
> When building a crosscompiling GCC 8.3.0 for powerpc the following "fix"
> is applied:
>
> $ diff /opt/pluto-targets/powerpc-rootfs/usr/include/open62541.h
> /opt/plutotoolchain/lib/gcc/powerpc-marel-linux-gnu/8.3.0/include-fixed/open62541.h
> 0a1,9
> > /* DO NOT EDIT THIS FILE.
> >
> > It has been auto-edited by fixincludes from:
> >
> > "/opt/pluto-targets/powerpc-rootfs/usr/include/open62541.h"
> >
> > This had to be done to correct non-standard usages in the
> > original, manufacturer supplied header file. */
> >
> 115c124
> < # ifndef __COUNTER__ /* PPC GCC fix */
> ---
> > # ifndef __COUNTER__ /* __PPC__ GCC fix */
>
> This fix is of course useless since it's editing a string inside a
> comment. Unfortunately it's also harmful since it causes this header
> file to be included in the compiler distribution in a way that it gets
> used before the system header. Then I went and upgraded the open62541
> library which made incompatible changes to its API (from v0.3 to v1.0 if
> it matters) causing builds to fail due to the wrong header being used.
>
> I can apply some hack to fix this locally in our packaging system so
> it's not an urgent error, and it would also go away by building the
> compiler with the new header installed. But I feel like those are
> workarounds that are not very future-proof.
>
> Is this something you think can/should be fixed? Do you need more info
> from me to figure out where and why this substitution happens?
It'll be a little while before I can re-pull all of GCC and look at it.
There's a substitution rule somewhere that effectively:
sed 's/PPC/__PPC__/'
somewhere. Add a constraint that prevents the application of that fix
when it appears after a "/*" then re-generate the C code that does all
the fixing.
Cheers - Bruce
> (I've sent this mail directly to you only instead of the mailing list
> since I don't know if I'm including enough information.)
Probably better to include gcc-patches as I have become less reliable on
email since I was retired.
