| Summary: | TLVs does not accept valid useBean action elements | ||
|---|---|---|---|
| Product: | Taglibs | Reporter: | Hans Bergsten <hans> |
| Component: | Standard Taglib | Assignee: | Tomcat Developers Mailing List <dev> |
| Status: | CLOSED FIXED | ||
| Severity: | blocker | ||
| Priority: | P3 | ||
| Version: | unspecified | ||
| Target Milestone: | --- | ||
| Hardware: | All | ||
| OS: | All | ||
Thanks for the report, Hans. I have fixed the error, although I implemented
a different check than the one you suggest. The validator wouldn't work
properly if I added the check you suggest: startElement() may legitimately be
concerned with elements that are not of the taglib's prefix. For instance,
it may desire to block ALL content within certain tags. (Consider
<c:choose>
<form>
...
)
You may have misunderstood the check for <jsp:text>: the goal isn't to rule
out and ignore all text in the page, but as the comment says, to
avoid "distinguishing between it and its characters." That is, I wouldn't
want a text element that contained nothing but whitespace, as --
<jsp:text> </jsp:text>
-- to invalidate a <c:choose> tag. The easiest way to do this is to ignore
<jsp:text> and </jsp:text>, and instead logically merge them with the characters
they contain (or oblivion if they contain no characters), given our constraints.
Thanks again.
|
The TLVs startElement() method goes through all its tests no matter which tag library an element belongs to. This results in the hasDanglingScope() method returning true for a perfectly valid <jsp:useBean> action. This is the case for the TLVs for all JSTL libraries. The solution is to add this right after the getLocalPart() call: // Just return if its an element with a foreign prefix if (!qn.startsWith(prefix + ":")) { return; } With this addition, the JSP_TEXT test can likely be removed.