Edward Hartnett

Here's some address santizer output: ``` ed@mikado:~/netcdf-c/nc_test4$ make tst_parallel6 && mpiexec -n 4 ./tst_parallel6 make: 'tst_parallel6' is up to date. *** Testing parallel I/O. *** testing a reported enddef problem...=================================================================...

And if I run this program with the new mode flag: NC_NODIMSCALE_ATTACH, the error does not occur. @dqwu you may wish to turn off dimscales in files with many variables....

For more info see: https://www.gnu.org/software/autoconf/manual/autoconf-2.67/html_node/Particular-Headers.html

It apparently requires a more recent version of autoconf as well. I notice that autoupdate changed the version of autoconf required at the top of the configure.ac. Since only developers...

What are you trying to do with the sample code? You have dims "lev" and "len" and you create variables "lev" and "len" - but you don't expect them to...

Yes, but I am right in the middle of major surgery on the NOAA GRIB2 libraries, so give me some time...

I actually got pretty confused looking at this PR. I don't understand it at all!

OK, I've studied this a bit more and it make sense. @wkliao have you measured any performance difference with these changes?

@WardF I am doing a bunch of extra testing before your release. It would be great if this change was merged, so it could be included in all my extra...

@WardF if this is going to be merged before 4.9.0, could you merge it please, so I can do some extra testing?