Le mercredi 20 juin 2007, Brian E. Fox a écrit : > I discovered http://jira.codehaus.org/browse/MNG-2961 while working on > mdep a while back. same for me while working on http://jira.codehaus.org/browse/MANTTASKS-18
> It should be an easy fix but I'm pondering what the > least intrusive fix is. thought it would be easy too, but as you mention, it must be the least intrusive: that's the hard part. > Clearly calling a Boolean method shouldn't > result in a change to some other variable so we should do something. > > > > Changing it to always returns xxx-SNAPSHOT as it does after calling > isSnapshot has the potential to break things depending on existing > behavior. The cleanest way forward is most likely to not have isSnapshot > modify the version, and to create a new method that returns xxx-SNAPSHOT > and leave getBaseVersion to return just the xxx-timestamp. (not sure > what this method would be called...getNonResolvedBaseVersion) > > > > WDYT? creating a new method is intrusive since this new method has to be used now: in this case, I know of o.a.m.artifact.repository.layout.DefaultRepositoryLayout.pathOf(), but I suppose there are others. To avoid a change in the API, I was thinking at modifying setVersion() and setBaseVersion() to do the job actually done in isSnapshot(). But I don't really understand the semantics of setBaseVersion(): as you noted, this has the potential to break things. To make a good decision, I think the first thing is to clearly describe the semantics of version and baseVersion: perhaps it is already clearly explained somewhere, but I don't know where. Depending the result, we'll be able to decide if the API has to be extended, or only the implementation fixed (and assume if it breaks something that works now but shouldn't because baseVersion had been used where version should have). --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
