Re: [RFC] CVS directory structure

On 15 Nov 2003 at 11:02, Xavier Noguer wrote: > I'm -1 on this. CVS is a tool for source control. Doing development > directly > on your CVS dir is a bad idea. > > Besides that, the new structure would not be the same as the > installed one > anyway. tests and data are in the first level in the installed pear > dir structure while inside each package's dir in the CVS one. Agreed. But when you run an example from "docs/examples" and if the "new" structure is used you could simply add e.g. Image_Text to your include_path and you're done. In my eyes CVS is not only used for development in this case but also for testing out a current version of a class. Therefor it's nice to be able to grab a current copy from CVS and using a simple modification to your ini_path you can try out the new version. I don't see why using the CVS-checkout-dir for development is bad. I'm doing it myself this way, many others do also, many tools do it this way (as Tobi has already mentioned on the list). You simply work in your source and when all files have been modified e.g. to contain a new function you simply check in that directory-tree. This way you are sure that you will check in all files that you modified and can asure that everything that's in CVS will function as you tested it locally. The way Bertrand (I think it was him - but that's not too important here) proposed seems a bit strange to me. Why should I do a checkout from CVS to the local disk, copy those files again somewhere, do modifications to e.g. 5 files and be sure to copy all 5 files back to check them in? If I see I'm stuck with the modifications I've done to those 5 files I can simply remove them, do a "cvs update" and have all my correct working-files I had before the "experiment". What's bad about it? And in connection with this the new structure in CVS might help ... Regards, Stefan