I'm pretty ok with stand along top-level fetch depends. My proposal is ortogonal. When any prefetch is strarted then: * nothing is done if md5sum corresponds to local file * file is downloaded === 1 time only === if md5sum mismatch or file just does not exist
> If some dependences is failed to get for the bad network status, > we had to wait for the long time building. If the dependency is not resoved (say right after you checkout the trunk from svs) the slow network will be an issues anyway. So let's have this simple but important fature for gulity search commit automation... Thanks Vladimir 2008/2/22, Sean Qiu <[EMAIL PROTECTED]>: > > There is a minor disadantage of this approach. > If some dependences is failed to get for the bad network status, we had to > wait for the long time building. > So IMHO, add a stand along top-level fetch-depends is better solution. > > > 2008/2/20, Alexey Petrenko <[EMAIL PROTECTED]>: > > > > 2008/2/19, Mark Hindess <[EMAIL PROTECTED]>: > > > > > > Thanks for doing this Nathan. > > > > > > Another thought, it would be useful to have a top-level fetch-depends > > > target too. It would need to call the three (?) more specific > fetch-* > > > targets. > > I'm not saying that top-level fetch-depends is bad idea but you can > > use -Dauto.fetch=true for this. > > > > SY, Alexey > > > > > > -- > Sean Qiu > http://xiaoxia.turendui.com >
