[Bug libstdc++/52447] Application crashes on SLES11 when using libstdc++.so.6.0.10 but runs fine with 6.0.8

[bkoz at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52447

Benjamin Kosnik <bkoz at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |bkoz at gcc dot gnu.org

--- Comment #7 from Benjamin Kosnik <bkoz at gcc dot gnu.org> 2012-04-25 02:35:21 UTC ---

Hmm. Reviewing this bug report, I see static initialization issues in large C++
projects, with many dependencies, involving three older and unsupported (from
the FSF perspective) toolchains, and no testcase.

To review:
gcc-3.4 ish = 6.0.3
gcc-4.1 ish = 6.0.8
gcc-4.3 ish = 6.0.10

Static locals got reworked at 4.0, so if you can run w/o error on 6.0.8 it's
probably not that. (Behaviour can be reverted w/ -fno-threadsafe-statics if you
are feeling like testing this.)

Without additional input, we're not going to be able to help you resolve this,
sorry. 

That might be ok for you, if  this was just a general "is this supposed to
work/are there known bugs" query. You have an answer.

If you are trying to get to the bottom of this, here are some things that might
be useful:

1) a way to reproduce, using one of the toolchain releases that is marked as
active on the gcc web site. Without this, I'm afraid we'll have to close this
bug report, and intend to do so after May 1, 2012.

2) porting to gcc-4.4/libstdc++.so.6.0.11+, instead of gcc-4.3. Gcc-4.4 was a
more widely deployed gcc release, and it is likely that integration issues with
the older (at this point, quite old) system environment may be better
understood. Those packages might be more polished, so maybe you'll just get
lucky.

You might want to see if SLES 11 has an alternate/newer compiler packaged
directly by the vendor, or use the opensuse gcc44 package. There may be a
subtle integration issue unknown to you.