On Mon, 20 Oct 2014, Harshad Sahasrabudhe wrote:

> Unfortunately, I'm using a complicated build which uses libMesh linked with
> a real PETSc. My code is then linked with libMesh, a real PETSc build and a
> complex PETSc build. It would be difficult to replicate the problem in an
> example.

Really?  I'd expect it would be difficult to replicate that build at
all.  I've seen the linker behavior "scream and die at link time when
a duplicate symbol definition exists", and I've seen the linker
behavior "silently choose the first definition encountered when a
duplicate symbol definition exists", but I've never seen the linker
behavior "correctly use whichever inconsistent version of a duplicate
symbol is appropriate for each time it is called".

If it *is* possible to link multiple versions of PETSc together in the
same executable, though, I'd love to do that officially.  It would be
great to be able to instantiate Real-valued or Complex-valued data
structures as requested at run time without having to choose only one
or the other at compile time.
---
Roy

------------------------------------------------------------------------------
Comprehensive Server Monitoring with Site24x7.
Monitor 10 servers for $9/Month.
Get alerted through email, SMS, voice calls or mobile push notifications.
Take corrective actions from your mobile device.
http://p.sf.net/sfu/Zoho
_______________________________________________
Libmesh-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-users

Reply via email to