Bugzilla – Full Text Bug Listing
| Summary: | URIUtil#toURI(URL) mangles fragments in file URLs | ||||||
|---|---|---|---|---|---|---|---|
| Product: | [Eclipse Project] Equinox | Reporter: | Markus Keller <markus.kell.r> | ||||
| Component: | Components | Assignee: | equinox.components-inbox <equinox.components-inbox> | ||||
| Status: | CLOSED DUPLICATE | QA Contact: | |||||
| Severity: | normal | ||||||
| Priority: | P3 | CC: | daniel_megert, john.arthorne, tjwatson | ||||
| Version: | 3.7 | ||||||
| Target Milestone: | --- | ||||||
| Hardware: | PC | ||||||
| OS: | Windows 7 | ||||||
| Whiteboard: | |||||||
| Bug Depends on: | 339422 | ||||||
| Bug Blocks: | |||||||
| Attachments: |
|
||||||
I'm not so sure any more what URIUtil#toURI(URL) is actually supposed to do, see bug 339422. Depending on that bug, the current implementation may actually be considered correct. Because java.net.URL and java.io.File#toURL are so broken, I'm not sure we can do much here. Take for example these two snippets:
new java.io.File("C:\\temp\\test\\a#b\\test.html").toURL().toString()
>> "file:/C:/temp/test/a#b/test.html"
new java.io.File("C:\\temp\\test\\test.html#anchor").toURL().toString()
>> "file:/C:/temp/test/test.html#anchor"
The first one is a directory containing the character #, the second is an anchor within a file. The resulting output is ambiguous. The URL class itself even parses them the same way:
new java.io.File("C:\\temp\\test\\a#b\\test.html").toURL().getRef()
>> "b/test.html"
new java.io.File("C:\\temp\\test\\test.html#anchor").toURL().getRef()
>> "anchor"
I.e., it always interprets the hash as the fragment separator, even though it might just be a path character. The fix from Markus will fix the second case, but break the first case.
I'm going to mark this as a duplicate of the more general bug 339422. *** This bug has been marked as a duplicate of bug 339422 *** |
Created attachment 189218 [details] Fix for one of the URIUtils I20110215-0800 URIUtil#toURI(URL) mangles fragments in file URLs (it wrongly encodes the # as %23 and also wrongly encodes other characters in the fragment). The method probably causes similar problems for the query part of a URL. I found the problem when trying to open generated Javadoc for a method in an external browser like this: - Preferences > General > Web Browser: Select "Use external web browser" - Generate Javadoc for a method, e.g. package p; public class C { public static void main(String[] args) { int i= 2; System.out.println(i); } } - Project > Generate Javadoc... (accept all defaults) - select "main" in the Editor and invoke Navigate > Open Attached Javadoc