On Wed, Mar 12, 2003 at 09:23:57AM +0100, Lars Gullik Bj?nnes wrote: > | This means I get to do one patch a day ... CVS is just useless at this. > > IMHO cvs has little to do with this.
It's got everything to do with it. The existence of cvs commit-locally would make everyone happy. > | *nobody* will read. Unless you know of some really smart way to get CVS > | to auto-switch between local and remote repositories: > > Then you need to make a script that traverses all the CVS dirs in your > checked-out sources. There it must change the Root file to point to > your local/remote repository. I belive gcc has that kind of tool as > part of their cvs tree. (contrib/newcvsroot) That's not "really smart". In fact, that's completely awkward and useless. > Do as I do: work on several trees at once. Are you kidding ? I have 10+ trees. They form a significant percentage of my disk space. Let's see: ideally, the patches committed recently would have been in about 30 different parts. Seeing as most are dependent on each other, I make that to be say 25 trees. Do you have 25 lyx trees ? Or perhaps you mean that you have a couple of trees with local CVS set up. That's not much better - you end up having to do all sorts of cross-tree diffs and management just to get a series of patches going. What I need is a way to commit a series of patches, one by one, locally, and *then* play them back in diff form. Trying to do that with CVS is painfully difficult. And I don't remember seeing you doing anything like this otherwise, which is surprising if it's as easy as you say. All I see are non-interdependent patches which are easy enough to manage, just by physically splitting up a diff or whatever. > For your first patch the one with the functions renameing and header > work, I would probably have been working in 3-4 different trees. > > It is not as if the changes where inter-dependant. Huh ? They were almost *ALL* in the same area of code i.e. conflicted head on. Patch C wouldn't apply or be possible without patches B and A, etc. regards john
