Hi, Is it not a decision that "optional" Tasks must be in same directrory that ant.jar (just like optional.jar) ? And so that is completely normal thet xdoclet.jar must be there as well. How XDoclet Task do to find other classes is up to itself. Regards Vincent
----- Original Message ----- From: "Hani Suleiman" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Monday, April 29, 2002 2:54 AM Subject: [Xdoclet-devel] FW: [Xdoclet-user] Xdoclet cvs issues > Forwarded since you guys probably care more about this that us end users... > > ------ Forwarded Message > From: Hani Suleiman <[EMAIL PROTECTED]> > Date: Mon, 29 Apr 2002 01:49:36 +0100 > To: xdoclet-user <[EMAIL PROTECTED]> > Subject: Re: [Xdoclet-user] Xdoclet cvs issues > > On 27/4/02 3:42 pm, "Aslak Hellesoy" <[EMAIL PROTECTED]> wrote: > > > -So what the code in xdoclet.jar is doing is to figure out where on the > > filesystem it is located, and then look for modules in the same directory. > > The reason why we have to construct a large classpath for ant in the bat/sh > > scripts is that this self-location-detection code in getXdocletJar() seems > > to return the location of ant.jar if we don't. I have no idea why it is this > > way, and I discovered that by putting it on the ant classpath the problem > > disappeared. We'd like to avoid this of course. > > > Ok, I looked into this further, and the bug is still present in the latest > CVS version of ant. > > The correct approach is for ant's classloader to extend SecureClassLoader, > and specify the correct defineClass method with the right CodeSource. > > <rant> > I'm surprised and disgusted by the ugliness of ant's classloader. Using > reflection to keep 1.1 compatibility is particularly stupid and ugly, and it > also means that now it's that much harder to fix the codesource bug given > the the brain damaged way the whole protectiondomain stuff is currently done > in AntClassLoader. A more sensible approach would be to define two > classloaders, a 1.1 and a 1.2 one, with a helper delegate class to hold > common functionality. > > Ant is a great tool, it's hard to find someone to argue against that. But as > usual with jakarta projects, the implementation is...interesting, to put it > politely. I was a much happier person living in ignorance and thinking that > ant is one of the few jakarta apps that is done right with decent code > behind it! > </rant> > > I hope someone with commit access to ant (or someone who has some kind of > influence on ant committers) can take this issue up with them, and fix this > so xdoclet doesn't end up suffering in result! > > Hani > > > > ------ End of Forwarded Message > > > _______________________________________________ > Xdoclet-devel mailing list > [EMAIL PROTECTED] > https://lists.sourceforge.net/lists/listinfo/xdoclet-devel > > _______________________________________________ Xdoclet-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/xdoclet-devel
