Le lundi 26 novembre 2007, Xavier Hanin a écrit : > On Nov 26, 2007 2:21 PM, Nicolas Lalevée <[EMAIL PROTECTED]> > > wrote: > > Le lundi 26 novembre 2007, Xavier Hanin a écrit: > > > On Nov 26, 2007 10:49 AM, Nicolas Lalevée > > > <[EMAIL PROTECTED]> > > > > > > wrote: > > > > Hi, > > > > > > > > I did an ivy plugin for Hudson, so Hudson will be able to trigger the > > > > build > > > > regarding of the dependencies declared in the ivy.xml. > > > > > > > > I did an implementation that does the similar to the ivy:buildlist > > > > ant task. > > > > Then I saw IVYDE-43 [1], which will quite do a similar job. And the > > > > solution > > > > exposed over there is to use a custom resolver. > > > > > > > > With the ModuleDescriptorParser, there are some "resolving", as we > > > > can get the > > > > full dependency graph. Ivy has to search for the ivy.xml the > > > > dependencies. > > > > > > > > I am quite confused by the API, there is simply no javadoc... > > > > > > > > So what should I use ? > > > > > > I think using something similar to how the buildlist works is fine. > > > With the ModuleDescriptorParser there is no resolving, ie no access to > > > a repository, it's only parsing a file to get the metadata. Then you > > > can > > > > ask > > > > > if a module is depending on another one based on the metadata, but it > > > > works > > > > > only for direct dependencies, not transitive ones. With a custom > > > > resolver > > > > > you can be aware whenever a resolution occurs and change the resolution > > > process, which is useful in IVYDE-43 case to replace the dependency on > > > > the > > > > > module jar(s) by a project dependency in Eclipse. But for what you do > > > in hudson using the parser is enough and simpler. > > > > ok. Thank you very much for the explanation ! > > > > > BTW, I will try to add some javadoc to some classes to help about that > > > > in > > > > > the future, could you tell me which classes/methods you looked at which > > > where the most confusing? > > > > For this particular issue, I was looking at ModuleDescriptorParser. But I > > did > > already looked at other parts of ivy. And what was disapointing me is > > that there were no javadoc on the Java interfaces. > > Indeed, Ivy code is not a very good example of how things should be done, > especially in the javadoc area. This is entirely my fault, when I started > writing Ivy in 2004 I was in a hurry and Ivy was just an internal tool for > my company... But this bad practice tends to follow the project for a long > time, we all learn with experience! Fortunately we are trying to add > javadoc when we maintain the code so it's slowly getting cleaner and > better. > > I wrote some javadoc from my > > > investigation, but unfortunatly I accidently removed my checkouted svn. > > Too sad, did you try restoring it with a disk rescue utility?
nope. My contribution was too poor (3 or 4 interfaces) and most probably not exact to take time to retreive it. But the next time will look into ivy code, I will repopulate what I found incomplete ;) Nicolas
