On Wed, May 27, 2015 at 9:59 AM, David Edelsohn <dje....@gmail.com> wrote: > On Wed, May 27, 2015 at 12:18 PM, H.J. Lu <hjl.to...@gmail.com> wrote: >> On Wed, May 27, 2015 at 9:05 AM, Peter Bergner <berg...@vnet.ibm.com> wrote: >>> On Wed, 2015-05-27 at 08:36 -0700, H.J. Lu wrote: >>>> On Wed, May 27, 2015 at 8:24 AM, Peter Bergner <berg...@vnet.ibm.com> >>>> wrote: >>>> > On Tue, 2015-05-26 at 16:40 -0500, Bill Schmidt wrote: >>>> >> Ah, never mind. I guess I need to run automake first. >>>> > >>>> > I ran the patch on powerpc64-linux (ie, Big Endian) both with and >>>> > without --enable-default-pie. Both bootstraps completed with no >>>> > errors and the without --enable-default-pie regtested without any >>>> > regressions. >>>> > >>>> > The --enable-default-pie regtesting shows massive failures that I >>>> > have to look into. I'm haven't determined yet whether these are >>>> > all -m32 FAILs or -m64 FAILS or both. I'll report back with more >>>> > info after I dig into some of the failures. >>>> >>>> Does --enable-default-pie work on powerpc64-linux? Do you >>>> get working PIE by default? Some GCC tests expect non-PIE. >>>> I fixed some of them: >>> >>> I haven't looked into any of the failures yet. That said, >>> powerpc64-linux is PIC by default, so I thought maybe PIE >> >> PIC != PIE. Is PIE the default for powerpc64-linux? Please >> show me >> >> # readelf -h /bin/ls >> >> on powerpc64-linux. >> >>> would just work. Maybe it does and it's just powerpc-linux >>> tests that are failing (I run the testsuite with both >>> -m32 and -m64). I won't know until I get some time to have >>> a deeper look. That said, if there is something you know >>> of that I should look for or at, I'd appreciate it. >>> >> >> You should first verify if --enable-default-pie generates a GCC which >> can produce a simple hello program. > > The --enable-default-pie patch itself works correctly -- as intended > -- for powerpc64-linux and powerpc64le-linux and the rs6000 parts of > the patch are okay with me. In other words, the option is fine but is > not enabled by default so it will not cause any problems.
Good to know. > Debugging the testsuite failures is a separate issue. Thanks. -- H.J.
