<div class="Ih2E3d">PM wrote:<br><div style="margin-left: 40px;">The biggest problem with this approach is that "diff" is much<br>
too coarsely grained to help us locate problems. For example, if<br>
we simply change a skin feature, HTML header, or even just the amount<br>
of whitespace that occurs in certain elements, then the above would<br>
report that every test has failed because it's no longer _exactly_<br>
the same as our standards in html.d/ . In other words, the "failures"<br>
it reports aren't really failures in PmWiki; each of the pages are<br>
still semantically correct -- they're still rendered properly in<br>
a browser -- but the diff command falsely reports them as having<br>
failed.<br></div>
<br></div>Given that wget produces an HTML file - and particularly if
it is an XHTML file - it is worth saying that it is possible to be more
subtle. <br><br>Of course we can use diff options that ignore
whitespace. We can also pretty-format the HTML and then diff that
canonical form (which ignores changes in cosmetic line breaks in the
original HTML). More usefully, we can choose to extract only specific
sub-trees, elements and/or attributes of interest from new and
reference HTML files (for example, using a simple XSLT report), and
then compare (only) those features. <br>
<br>Of course, the more thorough this kind of automated testing, the
more costly to set up. But even the simplistic diff is valuable - and
the gradual use of more sophisticated comparisons can significantly
reduce the number of false positives, while proving resilient over time.<br>
<br>Regards<br><font color="#888888"><br>Nigel Thomas<br><a href="http://www.preferisco.com/" target="_blank">http://www.preferisco.com</a></font><br><br><br>