Ed Hartnett

Can you build with --enable-logging, and then call nc_set_log_level(5) before running those commands? I have tried to build NCO here but could not get it to work.

But the names of levels are just integers ultimately, right? Can you show me an example of using your DAP logging in the code so I can take a look?...

@gsjaardema can you expand on why you would like this feature? What uses to you foresee? HPC applications are (ironically) often written and debugged without debuggers. (I use gdb, but...

@gsjaardema can I ask in general how you do I/O now from 100K+ cores? You use pnetcdf of course, but do you write directly from each core? Or do you...

Extra HPC featues would only be on for parallel processing systems. I'm confident I can map the other loggers to your system of warn, info, etc. When the time comes...

Maybe the trace system (nc_set_log_level() and the LOG(()) macro) should remain separate from the error logging mechanism? Or do we regard all trace statements as different levels of INFO log...

Here are all the LOG 0 statements in the code. Not many at all, compared to the total number of times we return an error code: ``` ed@mikado:~/tmp/netcdf-c$ find ....

The more I consider it, the more I think that tracing and error logging are two different things which should not be combined. However, each can be improved: the trace...

Why not continue to use the current LOG() mechanism for tracing. (We can change it to TRACE() to make it more clear.) It does not need to call the error...

BTW I have a massive refactor of libsrc4/libhdf5 under way to do lazy read of vars. So I would like to apply these changes myself, after the lazy read work...