Please review a fix to provide a new build-time tool to fix up the
output generated by the "pandoc" tool, which generates output that fails
to meet all of our documentation standards. Post-processing the output
was deemed a better solution than trying to modify the tool itself.
Most of the work is in the code for the new tool, written by me. The
remainder of the work, to integrate the tool into the build, was
contributed by Erik Joelsson.
The new tool comes in two parts. The first part uses an HTML parser to
stream the content of a file and to update it on the fly; the second
part of the file is just a copy-paste of a simple HTML parser that has
been used in other similar contexts, like the CodeTools DocCheck utility
[DocCheck].
The following changes are made to files processed by the tool, in order
to pass various accessibility checking tools.
* the <html> element is modified to set `lang="en"` and to remove
references to XML
* <main>...</main> is inserted around palpable content not already
included in a landmark region
* in tables, `scope="row"` is added to the cells in the column that
best define the row
* (minor) in the `<meta name="generator">` element, the content is
updated to indicate that the file has been modified by this tool
-- Jon
JBS: https://bugs.openjdk.java.net/browse/JDK-8224257
Webrev: http://cr.openjdk.java.net/~jjg/8224257/webrev.00/webrev/index.html
DocCheck: https://openjdk.java.net/projects/code-tools/doccheck/
