The "gnulib" package contains a number of floating-point test cases
that do not work correctly with the PowerPC long double ("double double")
format. These tests "accidentally" succeeded anyway in the big-endian
variant, but are now actually failing in little-endian mode.

As is usual for gnulib, those tests end up copied into the source code
of various packages that use gnulib, including coreutils, findutils,
grep, and libunistring.

A patch to fix the tests for ppc64le has been submitted to upstream
to the bug-gnulib mailing list. We'll work with upstream of the
other affected packages to make sure the copies are refreshed.

- grep-gnulib-ppc64le.patch: Fix imported gnulib long double math
tests for little-endian PowerPC.