Re: [PING][PATCH][1-3] New configure options that make the compiler use -fPIE and -pie as default option

[Magnus Granberg <zorry at gentoo dot org>]

söndag 11 januari 2015 14.29.56 skrev  Magnus Granberg:
> söndag 11 januari 2015 03.37.14 skrev  H.J. Lu:
> > On Sun, Jan 11, 2015 at 1:05 AM, Daniel Micay <danielmicay@gmail.com> 
wrote:
> > > On 11/01/15 03:52 AM, Daniel Micay wrote:
> > >> This branch works well for me. It doesn't automatically turn off the
> > >> linker flag when -fno-PIE or -fno-PIC are passed like most of the
> > >> wrapper scripts / downstream patches do, but that might not be a
> > >> problem
> > >> in practice.
> >  
> >  -fno-PIE and -fno-PIC aren't linking options.  We don't turn on PIE with
> > 
> > # gcc -fPIE ....
> > 
> > You need to use
> > 
> > # gcc -pie -fPIE
> > 
> > To turn off PIE, you should use
> > 
> > # gcc -no-pie -fno-PIE
> > 
> > > Er, Thunderbird did a terrible job handling my manual reply (wasn't
> > > subscribed to the list). Sorry about that.
> > > 
> > > For clarity, this is about H.J. Lu's branch at:
> > > 
> > > https://gcc.gnu.org/git/?p=gcc.git;a=shortlog;h=refs/heads/hjl/pie
> 
> It works fine for me but 6 test fails in the testsuite.
> Do we realy want to pass -fPIE -fPIC when someone do -fno-PIC or -fno-pic?
On the crtstuff do we and --fPIE or should the Makefile handel it?
I did let the Makefile handel it.
> 
> FAIL: gcc.dg/tree-ssa/ssa-store-ccp-3.c scan-tree-dump-times optimized
> "conststaticvariable" 1
> FAIL: gcc.target/i386/nop-mcount.c (test for excess errors)
> FAIL: gcc.target/i386/pr54445-2.c scan-assembler mov(b|zbl)[ \t]
> (%fs:)?tls_array@tpoff-1\\(%
> FAIL: g++.dg/other/anon5.C  -std=gnu++98 (test for excess errors)
> FAIL: g++.dg/other/anon5.C  -std=gnu++11 (test for excess errors)
> FAIL: g++.dg/other/anon5.C  -std=gnu++14 (test for excess errors)
> 
> /Magnus