------- Comment #19 from rguenther at suse dot de 2007-10-30 18:44 ------- Subject: Re: [4.3 Regression] typeinfo name referenced in ... defined in discarded section
On Tue, 30 Oct 2007, geoffk at geoffk dot org wrote: > On 30/10/2007, at 11:07 AM, rguenth at gcc dot gnu dot org wrote: > > > ------- Comment #16 from rguenth at gcc dot gnu dot org 2007-10-30 > > 18:07 ------- > > Geoff, your patch changed the typeinfo symbols for anonymous > > namespaces from > > having weak linkage to non-weak linkage. This caused the > > regression. Do you > > argue this change was deliberate and is the right thing? Why do you > > think so? > > Can you please restore the old behavior which is also that of other > > compilers > > for the linux targets? > > No, it changed the typeinfo symbols from being externally visible and > having weak linkage, to not being externally visible in which case the > weak-ness does not matter. > > The change is correct and necessary. You cannot access an anonymous > namespace from outside the translation unit it is defined in, so the > associated symbols do not need to be externally visible, and the > change is correct. It is necessary if you do not use random strings > in the symbol names, because otherwise the symbols will conflict. > > It does not matter what other compilers do, this is not an ABI issue. I see. So where is the bug here? We obviously see application builds fail because of this. Richard. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33871
