With the very helpful #3024 , we can now more easily approach the problem of what to do with the plugin path that the users specified in the configure step.
On library initialization, that path should be added to the plugin path.
This will cause everything to work out of the box for all users. Advanced users, thanks to #3024, will easily have the ability to take charge of the plugin path, and get full control. All other users will use the plugin path specified at configure time, and will not even have to know about it.
On HPC systems, sysadmins can build and install netCDF, and set whatever plugin directory they want, and users will get everything working out of the box without an environment variable, or even knowing anything about plugins or plugin paths.
This will reduce support requests to netCDF, because a plugin path problem will be perceived as a netCDF bug. Making it work out of the box without user involvement allows scientists to get right at their data without messing around with settings.
With the very helpful #3024 , we can now more easily approach the problem of what to do with the plugin path that the users specified in the configure step.
On library initialization, that path should be added to the plugin path.
This will cause everything to work out of the box for all users. Advanced users, thanks to #3024, will easily have the ability to take charge of the plugin path, and get full control. All other users will use the plugin path specified at configure time, and will not even have to know about it.
On HPC systems, sysadmins can build and install netCDF, and set whatever plugin directory they want, and users will get everything working out of the box without an environment variable, or even knowing anything about plugins or plugin paths.
This will reduce support requests to netCDF, because a plugin path problem will be perceived as a netCDF bug. Making it work out of the box without user involvement allows scientists to get right at their data without messing around with settings.