----- Original Message ----- From: "Peter Donald" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Thursday, April 26, 2001 11:46 PM Subject: Re: [VOTE] The late stuff
> At 09:26 26/4/01 -0400, Glenn McAllister wrote: > >Stefan Bodewig wrote: > > > >> Let's add another week for those two entries, I'll tally the votes on > >> Friday, May 4th. > >> > >> * Integration of the depends task and javac tasks > > > >After reading Stefan's and Jose's arguments, I'm -1 on this as well. It > >strikes me that <depends> is intended to make up for a lack in javac; if > >other compilers manage dependancy checking well on their own, why should > >we attempt to fix something that isn't necessarily broken with every > >invocation of the compiler by default? > > actually *no* compiler does it (reverse dependency checking) - most are > still broken but none of them intend to provide this functionality so lack > of this feature does not mean they are broken. > Yes, this is true. <depend> does more than Jikes does with its depend type features (+f option IIRC). When you compile a class, Jikes will make sure that any classes that it depends on are up to date. It does nothing about classes which depend on the class being compiled. Currently I run <depend> before every <javac> element. I don't find it that slow. So, I am +1 on integration, certainly making it optional. If it remains separate, and that seems likely, then perhaps it should be made a core task rather than being optional. It has no non-JDK dependencies. Conor
