[csg] HDF5 permissions problem/versions

[baumeier]

On our cluster running

LSB Version:	:core-4.1-amd64:core-4.1-noarch
Distributor ID:	CentOS
Description:	CentOS Linux release 7.8.2003 (Core)
Release:	7.8.2003
Codename:	Core

had trouble with the writing of the HDF5 file in xtp_map, for versions >=1.10 (maybe >=1.10.5)

xtp_map -t topol.tpr -c 770betaMADN.pdb -s 1betaMADN.xml -f system.hdf5 -v
==================================================
========   VOTCA (http://www.votca.org)   ========
==================================================
please submit bugs to bugs@votca.org
xtp_map, version 1.7-dev gitid: 2749fba (compiled Jun 10 2020, 10:28:58)
votca_csg, version 1.7-dev gitid: 2749fba (compiled Jun 10 2020, 10:25:33)
votca_tools, version 1.7-dev gitid: 2749fba (compiled Jun 10 2020, 10:25:11)
Reading file topol.tpr, VERSION 2019 (single precision)
Read MD topology from topol.tpr: Found 45430 atoms in 770 molecules.
The VOTCA pdb reader attempts to follow the official format:
http://www.wwpdb.org/documentation/file-format
NOTE the element symbol is important for associating a mass with
each atom, if no element symbol is specified the atom name will be
used and assumed to be an element symbol.
The pdb file format does not support pseudo atoms such as a CH3
and thus VOTCA cannot read such atoms in when using a .pdb file.
If you need to use pseudo atoms consider using a different file
format such as the lammps data format.
Read MD trajectory from 770betaMADN.pdb: found 1 frames, starting from frame 0
Frame with id 0 was not in statefile system.hdf5 ,adding it now.
an error occurred:
Could not access file system.hdf5 with permission to modify.

Success only possible with HDF5 1.8.21

xtp_map -t topol.tpr -c 770betaMADN.pdb -s 1betaMADN.xml -f system.hdf5 -v
==================================================
========   VOTCA (http://www.votca.org)   ========
==================================================
please submit bugs to bugs@votca.org
xtp_map, version 1.7-dev gitid: 2749fba (compiled Jun 10 2020, 11:28:26)
votca_csg, version 1.7-dev gitid: 2749fba (compiled Jun 10 2020, 11:23:24)
votca_tools, version 1.7-dev gitid: 2749fba (compiled Jun 10 2020, 11:22:10)
Reading file topol.tpr, VERSION 2019 (single precision)
Read MD topology from topol.tpr: Found 45430 atoms in 770 molecules.
The VOTCA pdb reader attempts to follow the official format:
http://www.wwpdb.org/documentation/file-format
NOTE the element symbol is important for associating a mass with
each atom, if no element symbol is specified the atom name will be
used and assumed to be an element symbol.
The pdb file format does not support pseudo atoms such as a CH3
and thus VOTCA cannot read such atoms in when using a .pdb file.
If you need to use pseudo atoms consider using a different file
format such as the lammps data format.
Read MD trajectory from 770betaMADN.pdb: found 1 frames, starting from frame 0
Frame with id 0 was not in statefile system.hdf5 ,adding it now.
Writing MD topology (step = 0, time = 0) to system.hdf5
... .
The VOTCA pdb reader attempts to follow the official format:
http://www.wwpdb.org/documentation/file-format
NOTE the element symbol is important for associating a mass with
each atom, if no element symbol is specified the atom name will be
used and assumed to be an element symbol.
The pdb file format does not support pseudo atoms such as a CH3
and thus VOTCA cannot read such atoms in when using a .pdb file.
If you need to use pseudo atoms consider using a different file
format such as the lammps data format.

However, there the PDB reader note appears twice. Is this expected behavior, and if so, could it point to what is the problem with the 1.10 version?

[JensWehner]

Concerning the pdb reading how many timesteps are in the pdb?

[baumeier]

Single frame only.

[junghans]

@JoshuaSBrown wrote the pdb reader, so he might know.

[JoshuaSBrown]

@baumeier can you send me the files you were working with? I will try to take a look at it tonight. Is this on master or stable?

[JoshuaSBrown]

On first glance though this looks like the problem is the installation of hdf5.

[JoshuaSBrown]

That message is simply a notification, repetition in the notification is not necessarily an indication of an error. It just means that you are opening the file on two separate occasions, though that should probably be fixed.

[baumeier]

This was master.
Do you know why the file is opened twice? At which stages?

[JoshuaSBrown]

No I think it is looking to see if it can read from the next frame, that message gets triggered because it doesn't check to see if the file stream is at the end of the file first, but it should not crash because it won't actually read the next frame if there is not one.

Also this stuff is in csgapplication which needs a complete refactor so beware. E.g. I'm not sure I can explain the logic of the below section that triggers this.

_traj_reader->FirstFrame(master->_top);
if (master->_top.getBoxType() == BoundaryCondition::typeOpen) {
  std::cout << "NOTE: You are using OpenBox boundary conditions. Check if this is intended.\n" << std::endl;
}                                                                                                                                                                                                              // seek first frame, let thread0 do that                                                                                                                                                                       bool bok;
for (bok = true; bok == true; bok = _traj_reader->NextFrame(master->_top)) {
  if ((has_begin && (master->_top.getTime() < begin)) || first_frame > 1) {                                                                                                                                        
    first_frame--;                                                                                                                                                                                                 
    continue;                                                                                                                                                                                                    
  }                                                                                                                                                                                                              
  break;                                                                                                                                                                                                       } 

[JensWehner]

I think we should just put the warning message somewhere else, maybe where the file is opened.

[JensWehner]

This issue should be split into a .pdb reader problem and the .hdf5 problem

[JoshuaSBrown]

I think we should just put the warning message somewhere else, maybe where the file is opened.

This is a good idea, I think that the lammps readers could also benefit from this. The other option is to have a static flag so that you only display the message once no matter once.