Greg Eisenhauer

Hi. We've recent discovered that this error pops up when the two different instances of ADIOS end up selecting different DataTransports in the SST data plane layer. A pull request...

So, do you think the issue is that it times out? Or that the suicide is not clean? The timeout is by design of course, because servers started via ssh...

Yeah, you're probably the first one to let the server timeout with coredumpsize set to unlimited. Generally on the HPC machines where the server runs coredumpsize is set to 0,...

It exits cleanly for me... You might have to do some debugging yourself to get us more info for this to be fixed. eisen@Endor build % lldb bin/remote_server (lldb) target...

It may be that it depends upon state left from some prior event. I.E. maybe if it times out while a client is still connected (but inactive), or something like...

adios2NPTypes.py is imported into other tests, so it doesn't need to be in CMakeLists.txt. The HDF5 one would have to be protected because we don't always build with HDF5 available...

Does it succeed? My only observation would be that if it uses the HDF5 engine, it needs to be inside a conditional in CMakeLists.txt so that it only tries to...

> Although I never touched the file, yes I think it should be included if ADIOS is compiled with HDF5 support. Ah, I hadn't looked at the file history. Well,...

I'd say it counts as a bug. This wasn't really an issue with BP3, but both BP4 and BP5 create directories of the appropriate name and just overwrite what's there,...

> I would also add a git commit hash of the build into BP4/BP4 files and into the command output of 'bpls -h'" in order to track faster what version...