TL;DR: Run update_unicode.sh after the publication of a new Unicode standard and commit the resulting unicode-widths.h file.

The long version

The Git source code ships the file unicode-widths.h which contains tables of zero and double width Unicode code points, respectively. These tables are generated using update_unicode.sh in this directory. update_unicode.sh itself uses a third-party tool, uniset, to query two Unicode data files for the interesting code points.

On first run, update_unicode.sh clones uniset from Github and builds it. This requires a current-ish version of autoconf (2.69 works per December 2016).

On each run, update_unicode.sh checks whether more recent Unicode data files are available from the Unicode consortium, and rebuilds the header unicode-widths.h with the new data. The new header can then be committed.