Hello every one,
I am trying to install abinit.9.4.2 in Ubuntu 20.04 (i7 lenovo laptop).
I followed what is recommended in Ubuntu - abinit
using the follwing settings: sam.ac9.log (1.9 KB)
at the end I get:
±-----------------------------------------------------------------+
| WARNING : NetCDF is not WORKING ! |
±-----------------------------------------------------------------+
| Please specify the installation directory for NetCDF |
±-----------------------------------------------------------------+
±-----------------------------------------------------------------+
| WARNING : NetCDF Fortran is not WORKING ! |
±-----------------------------------------------------------------+
| Please specify the installation directory for Netcdf-fortran |
±-----------------------------------------------------------------+
±-----------------------------------------------------------------+
| Problems must be solved before continuing |
±-----------------------------------------------------------------+
config.log (176.9 KB)
I attached the config.log as well for you kind consideration.
Many thanks,
Samir
Hi @samabdel! Is it possible to look at your .ac9 file used for the configure? It must be named appropriately with the result of the hostname command ({hostname}.ac9) or you must launch the configure script with the option --with-config-file=your_config_file.ac9. In this .ac9 file you must point out where the netcdf library is installed (you can use the command with_netcdf_fortran=$(nf-config --prefix).
Dear jmbgmx,
Thank you so much…well done! it’s working very well now.
In the build parameters (the attached file), I have:
HDF5 enabled : yes (MPI support: no)
NetCDF enabled : yes (MPI support: no)
NetCDF-F enabled : yes (MPI support: no)
Does it mean none of the HDF5, NetCDF, and NetCDF-F libraries is parallel?
I need the parallel HDF5 for another code (the FLAPW Fleur code) to be able to compare the pseudopotential (Abinit) to an all-electron (Fleur) code.
Many many thanks again,
A+
Samirbuild-parameters.log (1.1 KB)
Hi,
Apparently it is not what I thought…even the configuration is successfully finished, this is what I get when making make -j8:
sockets.c: In function ‘open_socket’:
sockets.c:84:23: error: storage size of ‘hints’ isn’t known
84 | struct addrinfo hints, *res;
| ^~~~~
sockets.c:90:24: error: ‘AI_PASSIVE’ undeclared (first use in this function)
90 | hints.ai_flags = AI_PASSIVE;
| ^~~~~~~~~~
sockets.c:90:24: note: each undeclared identifier is reported only once for each function it appears in
sockets.c:93:16: warning: implicit declaration of function ‘getaddrinfo’ [-Wimplicit-function-declaration]
93 | ai_err = getaddrinfo(host, service, &hints, &res);
| ^~~~~~~~~~~
sockets.c:97:26: error: dereferencing pointer to incomplete type ‘struct addrinfo’
97 | sockfd = socket(res->ai_family, res->ai_socktype, res->ai_protocol);
| ^~
sockets.c:103:7: warning: implicit declaration of function ‘freeaddrinfo’ [-Wimplicit-function-declaration]
103 | freeaddrinfo(res);
| ^~~~~~~~~~~~
make[5]: *** [Makefile:1224: sockets.o] Error 1
Thanks again,
Samir
Hi,
This is exactly what I suspected…It turns out that this error is due to the -std=c99 gcc compiling option.
This means that for some reason (and I don’t know which?) abinit does not need gcc -std=c99 to use c99.
Many many thanks again,
and hope that I will be able to install HDF5 parallel too…with your help.
Samir
Hi,
Apparently it is not over yet!!!
This is what I get when I run a parallel abinit:
ERROR
src_file: m_nctk.F90
src_line: 590
mpi_rank: 0
message: |
–netcdf-default is on but netcdf library does not support MPI-IO. Aborting now
#####################
Many thanks,
Samir
means that you want to change the default behaviour of Abinit so that all files are produced with netcdf without having to set iomode 3 in the input file.
This also means that Abinit will use netcdf to write/read the WFK.nc files and IO of these files with paral_kgb =1 requires MPI-IO capabilities.
This is the reason why the code is checking whether your netcdf library supports MPI-IO
in order to avoid runtime errors at end of a big calculation when we try to write the WFK in parallel.
If you don’t need to run big calculations with paral_kgb 1 and netcdf output, I would suggest to avoid using enable_netcdf_default=“yes”.
Hi samabdel,
If you still have problem related to netcdf or netcdf fortran. Then you should try this…
you can go build directory then after this you will go in fallbacks folder and run this command…
./build-abinit-fallbacks.sh
Then you will get path format of netcdf… after this you add this path in your configure file. Then you will go in build and configure this file with this command …/configure --with-config-file="conf. file"