This is the mail archive of the
mailing list for the GCC project.
Re: [PATCH] Fix PR86523
On Tue, 17 Jul 2018, Rainer Orth wrote:
> Hi Richard,
> > The following fixes PR86523, we failed to assing DIE parents to some
> > function-local entities with the idea scope vars would pick them up
> > but that's not true for some of them.
> > Bootstrapped and tested on x86_64-unknown-linux-gnu, LTO bootstrapped
> > on the GCC 8 branch (it's said LTO bootstrap is broken on trunk).
> the new testcases FAIL on Solaris:
> +FAIL: g++.dg/lto/pr86523-1 cp_lto_pr86523-1_0.o-cp_lto_pr86523-1_0.o link, -O2 -flto -g -shared
> +FAIL: g++.dg/lto/pr86523-2 cp_lto_pr86523-2_0.o-cp_lto_pr86523-2_0.o link, -O2 -flto -g -shared
> 32-bit sparc and 64-bit x86 only
> On sparc:
> ld: warning: unwind table: file /var/tmp//ccRLBN4b.ltrans0.ltrans.o: section .eh_frame: suspicious InitialLoc value 0: binary search table will be incomplete if section requires runtime relocation
> On x86:
> Text relocation remains referenced
> against symbol offset in file
> .text (section) 0x20 cp_lto_pr86523-1_0.o
> ld: fatal: relocations remain against allocatable but non-writable sections
> The latter happens because the Solaris linker defaults to -z text, and
> the tests aren't built as PIC. The following patch fixes this,
> requiring fpic and shared support along the way, too, and also fixes the
> Solaris/SPARC failure.
> Tested on i386-pc-solaris2.11, sparc-sun-solaris2.11, and
> x86_64-pc-linux-gnu. Ok for mainline?