> > I guess it depends on what would happen if you then passed the > transaction to rpm. If your "pre-rpm transaction" fails, and you then > pass it to rpm, will that fail too? If so, there's no point in doing > it. If you attempt to install an already installed pkg or upgrade an already current pkg - yes the rpm transaction will fail. > > Maybe we need a "--use-multiple-rpm-transactions" flag? Then internally > use multiple rpm transactions instead of a single one. > > But if we do that, and there's nothing to be gained over doing: > yum update pkg1 > yum update pkg2 > then I'd suggest just doing nothing. (I'm not saying there is nothing > to be gained, I'm hypothicating.) There's no sense having yum do the > looping internally unless some speed gain can be realized. I started > this thread because I thought yum could do a better job than me in > looping, perhaps by computing dependencies only in one pass, instead of > multiple ones. If that's not the case, then lets leave it like it is. You could create a new transaction for each but you would lose out in speed more or less. Right now from pressing enter to dep resolution is about 6s I think there are some places where I can trim that but even so - it's 6seconds. I'm hoping to work on yum some more tomorrow - get a 1.96 out w/all the bug fixes of this week and work on the checkfunc for downloading stuff maybe, if everyone is good I'll be able to get something together for 1.0.1 too. -sv
