Segmentation fault (core dumped) issue when Reading hdf5 files

[thegiantspaceman]

I am using linux OS and have installed NCL using the conda install method here. I have NCL version 6.6.2 and it works fine except for the fact that it cannot read hdf5 files. To illustrate this error, below is a test script to write and then read an hdf5 file. The script is able to write to hdf5, but once it goes to read hdf5 it gives "Segmentation fault (core dumped)." How can I fix this issue?

begin
fn = "./out.hdf5"
a= (/1,2/)
;WRITE
f = addfile(fn, "c")
f->a = (/a/)
;READ
f = addfile(fn, "r")
s = f->a
end

[Rick Brownrigg]

Hi, I don't know for certain, but I wonder if the file you wrote has not been properly closed at the time you attempt to read it? Perhaps add a "delete(f)" statement before opening the file for reading. This should delete the filevariable, not the file, and in doing so, I speculate it will close the file. Rick

…

On Tue, Jan 3, 2023 at 9:27 AM thegiantspaceman ***@***.***> wrote: I am using linux OS and have installed NCL using the conda install method here <https://www.ncl.ucar.edu/Download/conda.shtml>. I have NCL version 6.6.2 and it works fine except for the fact that it cannot read hdf5 files. To illustrate this error, below is a test script to write and then read an hdf5 file. The script is able to write to hdf5, but once it goes to read hdf5 it gives "Segmentation fault (core dumped)." How can I fix this issue? begin fn = "./out.hdf5" a= (/1,2/) ;WRITE f = addfile(fn, "c") f->a = (/a/) ;READ f = addfile(fn, "r") s = f->a end — Reply to this email directly, view it on GitHub <#196>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADLWOXUMZ2HRONYRM7T7ABTWQRHPHANCNFSM6AAAAAATP3R4NI> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[thegiantspaceman]

Thank you for the quick reply Rick. I added "delete(f)" before reading, but I still get the same seg fault issue. I also have tried to open an hdf5 file output from a model and I encounter the exact same error. Should I try to build NCL differently on my machine? Do you have any other suggestions?

-Reuben

[Rick Brownrigg]

Regrettably, I don't really know what else to suggest. There's the pre-compiled binary option, in addition to the conda installation ( https://www.ncl.ucar.edu/Download/). I would not recommend trying to build NCL from source unless you have a great deal of experience building out source code packages. Best of luck... Rick

…

On Tue, Jan 3, 2023 at 11:33 AM thegiantspaceman ***@***.***> wrote: Thank you for the quick reply Rick. I added "delete(f)" before reading, but I still get the same seg fault issue. I also have tried to open an hdf5 file output from a model and I encounter the exact same error. Should I try to build NCL differently on my machine? Do you have any other suggestions? -Reuben — Reply to this email directly, view it on GitHub <#196 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADLWOXREBIKPONO4WP2OVVLWQRWGPANCNFSM6AAAAAATP3R4NI> . You are receiving this because you commented.Message ID: ***@***.***>