This is an archived snapshot of W3C's public bugzilla bug tracker, decommissioned in April 2019. Please see the home page for more details.
In CSSUtils and HTTPXHTMLResource, when we encounter a URI, we get it to be parsed using new URI(), which sometimes throw a URISyntaxException. Currently, the only thing that happens where that exception is thrown is a log line that never gets seen by the end-user. There are 2 problems: * the URI construct throws URISyntaxException for URIs that most browsers don't have a problem dealing with; typically, if the "|" character in the URI. This means there are images that we don't check due to this limitation ; maybe we could %-encode the list of "unwise" characters in RFC2396 as part of the pre-processing to avoid that problem? * for URIs that we really can't parse (or if we don't implement the above), instead of using log.info(), we should report the URI parsing problem as a HTTP_RESPONSE-1 issue
Partially fixed -- I added new code to catch URISyntaxException where it occurs from the URI constructor, rewrite the URL escaping "unwise" characters, and try again.
I committed a hack-ish solution to this. It forms a valid but non-existent URL out of a badly formed URL, which induces a failure later. This is really a corner case so I am basically OK with a hacky solution. I think we don't need to spend a lot of effort complicating moki.xml and the code to handle this differently.