Global destructor order problems

[Michael Gmelin]

On Thu, 27 Jun 2013 17:04:28 +0300
Konstantin Belousov <kostikbel at gmail.com> wrote:

> On Thu, Jun 27, 2013 at 03:41:48PM +0200, Dimitry Andric wrote:
> > On 2013-06-27 01:56, Michael Gmelin wrote:
> > > On Thu, 27 Jun 2013 00:28:33 +0300
> > > Konstantin Belousov <kostikbel at gmail.com> wrote:
> > >
> > >> On Wed, Jun 26, 2013 at 11:17:41PM +0200, Michael Gmelin wrote:
> > >>> Are you both on the same architecture?
> > >>
> > >> I tested both on amd64 and i386. For i386, it was -m32 for
> > >> clang, and native 32bit gcc 4.8.1, stock build from the tarball.
> > >>
> > >
> > > For completeness sake I tested once more using various compilers
> > > including gcc 4.8.1 on 10-CURRENT amd64 (see below).
> > 
> > I have now tested the lang/gcc44, 46, 47, 48 and 49 ports, on both
> > i386 and amd64. All result in the incorrect destructor order.
> 
> My build of gcc is stock, not from the ports.  Another detail, I built
> the compiler on stable/9, but use it on both stable and current.  I am
> very curious what makes the my build to generate the code which uses
> PLT instead of direct reference, but have no time to investigate.
> 
> I think that we could revert the termination calls to the functions
> from the dso being unloaded, but this is quite unfortunate, since it
> will restore the endless series of 'segfault at the process
> termination' reports.

So now what?

-- 
Michael Gmelin