On 5/21/21, G. Branden Robinson <[email protected]> wrote: > The last is useful to catch almost anything that might go wrong in PS > or PDF output, without the tedium of visually comparing such output.
Two other possibilities for scriptably regression-testing PS/PDF output: * In cases, like man pages, that aren't doing a lot of tricky typesetting maneuvers (e.g., dropping the E in the middle of TEX), saving expected "groff -a" output and then diffing new runs against that often reveals when things change on the typeset page. * The diffpdf.py script (http://github.com/brechtm/rinohtype/blob/stable/tests_regression/helpers/diffpdf.py), which is an updated version of the script I posted about in January (http://lists.gnu.org/archive/html/groff/2021-01/msg00101.html), compares pixels on the page of two PDFs.
