I think the issue causing the test failure is likely to be related to the nonstandard long doubles used on powerpc processors.
See this MIPS report for a similar issue. Disabling long doubles on platforms with nonstandard implementations was suggested. https://gitlab.gnome.org/GNOME/goffice/-/issues/59 > On Aug 24, 2025, at 04:41, Jeroen Diederen <[email protected]> wrote: > > It is the same fix as applied for many years in ppc64le. This is a release > arch. > ttps://bugzilla.gnome.org/show_bug.cgi?id=748504 > > John Paul Adrian Glaubitz schreef op 2025-08-24 17:47: >> Hello Jeroen, >>> On Sun, 2025-08-24 at 16:44 +0700, Jeroen Diederen wrote: >>> The source package goffice does not build from source on ppc64. To make >>> it build I added ppc64 in debian/rules: >>> --- >>> override_dh_auto_test: >>> ## https://bugzilla.gnome.org/show_bug.cgi?id=748504 >>> ifeq ($(DEB_BUILD_ARCH),$(filter $(DEB_BUILD_ARCH),powerpc ppc64el >>> ppc64)) >>> @echo "======== will not abort on test(s) failure ========" >>> -dh_auto_test >>> else >>> dh_auto_test >>> endif >>> --- >>> The source is then built successfully on ppc64. >>> The binary libgoffice-0.10-dev from this source is important for the >>> build of gnumeric which couldn't be built on ppc64 for a very long time. >>> Please fix. >> Disabling the testsuite isn't really a fix, just a workaround which can cause >> the package to cause crashes on the end user system. A proper fix would be >> investigating and fixing why the testsuite is failing. >> Adrian >
