Problem: On very larger clusters command line operations of a large
number of nodes (e.g. node[1,3,5,...,4997,4999]) can exceed internal
buffers, leading to errors.
Increase the internal maximum line length from 8K to 128K, which should
handle any HPC cluster in existence today and larger ones in the future.
Fixes #193
side note, for the test I just created the giant input and put it in a file in t/data b/c I was not confident that hostlist generating tools that could generate it would exist on many environments. (i.e. reproducer I was shown used cluset which I don't think is that common. Neither would flux hostlist, etc.)
Problem: On very larger clusters command line operations of a large number of nodes (e.g. node[1,3,5,...,4997,4999]) can exceed internal buffers, leading to errors.
Increase the internal maximum line length from 8K to 128K, which should handle any HPC cluster in existence today and larger ones in the future.
Fixes #193
side note, for the test I just created the giant input and put it in a file in
t/data
b/c I was not confident that hostlist generating tools that could generate it would exist on many environments. (i.e. reproducer I was shown usedcluset
which I don't think is that common. Neither wouldflux hostlist
, etc.)