Charm++ parallel programming system

[Jozsef Bakosi <jbakosi AT lanl.gov>]

Hi folks,

I'd like some help debugging this because this prevents us from upgrading to
v6.8. The symptom is that I get the following error going from v6.7.2 to 
v6.8.0:

Runtime error:

$ ./<my-executable>
symbol lookup error: <charm-install-dir>/lib_so/libconv-core.so: undefined 
symbol: __executable_start

ldd output:

$ ldd -r <my_executable>
...
undefined symbol: __executable_start    
(<charm-install-dir>/lib_so/libconv-core.so)

Compile and link are fine. This is only an error using g++ on linux, v7, 6, 
and
5, and not with clang or intel compilers.

Can someone please point me to how I might be able to debug this?

Thanks,
Jozsef

On 02.05.2018 12:33, Phil Miller wrote:
> On Mon, Feb 5, 2018 at 11:15 AM, Jozsef Bakosi 
> <jbakosi AT lanl.gov>
>  wrote:
> 
> > That's correct. Fully recompiled, several times using different compilers
> > and on two machines, all third-party libraries (of which charm is one) and
> > the code itself, from scratch.
> 
> OK, I've got nothing off the top of my head, then. Let the list know, and
> one of the current staff will hopefully take a look.
>
> >
> > On 02.05.2018 11:13, Phil Miller wrote:
> > >    Just to make sure, you have fully recompiled your code with the
> > >    upgraded version of Charm++? The system offers no ABI preservation 
> > > from
> > >    one version to the next.
> > >
> > >    On Mon, Feb 5, 2018 at 11:08 AM, Jozsef Bakosi 
> > > <[1]jbakosi AT lanl.gov>
> > >    wrote:
> > >
> > >      Hi folks,
> > >
> > >      Updating from Charm++ v6.7.1 to 6.8.2 breaks executables with the
> > >      runtime error in the subject. Only with the gnu compiler suite,
> > >      consistently with version 5, 6, and 7. Intel and clang are fine.  I
> > >      wonder if this has anything to do with the following feature,
> > >      introduced in 6.8.0: "Charm++ programs can now define their own
> > >      main() function, rather than using a generated implementation from 
> > > a
> > >      mainmodule/mainchare combination.  This extends the existing
> > >      Charm++/MPI interoperation feature." Could some point me to how I 
> > > can
> > >      go about diagnosing and/or fixing this problem?
> > >      Thanks,
> > >      Jozsef