On Fri, Feb 01, 2019 at 02:12:00AM +0100, Luc Van Oostenryck wrote: > On Thu, Jan 31, 2019 at 06:44:13PM +0100, Uwe Kleine-König wrote: > > On 1/31/19 3:26 PM, Luc Van Oostenryck wrote: > > > > > > Does 'uname -m' or 'uname -s' returns 'gnu' on Hurd? > > > > ukleinek@exodar:~$ uname -s > > GNU > > ukleinek@exodar:~$ uname -m > > i686-AT386 > > OK, thanks. > > > >> and on x32 I got: > > >> > > >> /usr/bin/ld: compile-i386.o: relocation R_X86_64_32S against `.data' can not be used when making a PIE object; recompile with -fPIC > > >> /usr/bin/ld: final link failed: nonrepresentable section on output > > >> collect2: error: ld returned 1 exit status > > > > > > I didn't had this when testing on x32 but I see that here the compiler > > > is called here with: > > > -specs=/usr/share/dpkg/pie-link.specs > > > -Wl,-z,relro -Wl,-z,now > > > The first one is clearly part of the problem, maybe in combination > > > with the other two. > > > From where do they come? Why? It's not used for amd64 and other archs. > > > > The linker flags come in from the automatic hardening that is suggested > > to use for Debian packages. In the amd64 buildlog I see: > > > > ... > > gcc-8 -Wl,-z,relro -Wl,-z,now -Wl,--as-needed compile.o libsparse.a > > compile-i386.o -o compile > > ... > > > > not sure where the -specs=... part comes from. compile-i386 > > I think the cause may be that if PIE is wanted (via the -specs=...) > then -fPIC is also needed at compile time. > So, I guess that the -specs is given via LDFLAGS but not CFLAGS. More, exactly, if -specs=/usr/share/dpkg/pie-link.specs is given in LDFLAGS then /usr/share/dpkg/pie-compile.specs should also be given in CFLAGS but is not (or is it pie-compiler.specs?). Both should be set via "DEB_BUILD_MAINT_OPTIONS = hardening=+all" in debian/rules but it is said in dpkg-buildflags.1 that "buildflags.mk" should be included (and probably DPKG_EXPORT_BUILDFLAGS be set). Now, I don't understand why LDFLAGS should be set anyway but not CFLAGS. Best regards, -- Luc
