Add define for MIPS_XHASH

[LinuxUserGD]

Fixes undeclared identifier error on musl libc (Gentoo amd64/clang), see https://mail-index.netbsd.org/source-changes-hg/2021/05/01/msg290415.html

clang++ -std=c++17 -DPACKAGE_NAME=\"patchelf\" -DPACKAGE_TARNAME=\"patchelf\" -DPACKAGE_VERSION=\"0.17.2\" -DPACKAGE_STRING=\"patchelf\ 0.17.2\" -DPACKAGE_BUGREPORT=\"\" -DPACKAGE_URL=\"\" -DPACKAGE=\"patchelf\" -DVERSION=\"0.17.2\" -DHAVE_CXX17=1 -I.    -Wall -std=c++17 -D_FILE_OFFSET_BITS=64     -O3 -pipe -march=skylake -mtune=skylake -D_FORTIFY_SOURCE=3 -stdlib=libc++ -c -o patchelf.o patchelf.cc
patchelf.cc:1087:33: error: use of undeclared identifier 'DT_MIPS_XHASH'
            } else if (d_tag == DT_MIPS_XHASH) {
                                ^
1 error generated.
make[1]: *** [Makefile:380: patchelf.o] Error 1
make[1]: Leaving directory '/var/tmp/portage/dev-util/patchelf-0.17.2/work/patchelf-0.17.2/src'
make: *** [Makefile:448: all-recursive] Error 1

[tgbugs]

Is it safe to define that constant if it has not already been defined? Is it a glibc specific constant or is it an implementation of something that is standard on MIPS?

[LinuxUserGD]

Is it safe to define that constant if it has not already been defined? Is it a glibc specific constant or is it an implementation of something that is standard on MIPS?

It's a constant in binutils so probably not specific to glibc-only systems I think

Edit: Also seems to be defined in

patchelf/src/elf.h

Line 2011 in ea2fca7

#define DT_MIPS_XHASH 0x70000036

[brenoguim]

Edit: Also seems to be defined in

patchelf/src/elf.h

Line 2011 in ea2fca7

#define DT_MIPS_XHASH 0x70000036

Since this is already defined, why would this PR be necessary? Perhaps the real problem is another one?

[LinuxUserGD]

Edit: Also seems to be defined in

patchelf/src/elf.h

Line 2011 in ea2fca7

#define DT_MIPS_XHASH 0x70000036

Since this is already defined, why would this PR be necessary? Perhaps the real problem is another one?

Maybe elf.h from musl libc is included first https://git.musl-libc.org/cgit/musl/tree/include/elf.h which doesn't define DT_MIPS_XHASH so _ELF_H is already defined without constants from https://github.com/NixOS/patchelf/blob/master/src/elf.h

There's also a gentoo bug https://bugs.gentoo.org/860888 removing DT_MIPS_XHASH which might be a better approach.

[brenoguim]

I'm confused. So even though patchelf contains the elf.h file used to build and test, it needs to also work with some other elf.h? It seems backwards to me.
If the problem is indeed musl providing another elf.h (needs verification) it would be nicer if it just didn't unless we explicitly ask for it.
Let me try to reproduce the issue and better understand what is going on.

[LinuxUserGD]

I'm confused. So even though patchelf contains the elf.h file used to build and test, it needs to also work with some other elf.h? It seems backwards to me. If the problem is indeed musl providing another elf.h (needs verification) it would be nicer if it just didn't unless we explicitly ask for it. Let me try to reproduce the issue and better understand what is going on.

Ok, Gentoo actually removes removes the file which causes the build failure: https://gitweb.gentoo.org/repo/gentoo.git/tree/dev-util/patchelf/patchelf-0.17.2.ebuild#n17
Sry, seems like code should be fine upstream, I guess a downstream gentoo patch might be better then.