On Jun 6, 2012, at 3:48 PM, Jed Brown wrote:

On Wed, Jun 6, 2012 at 3:40 PM, Satish Balay <[email protected]> wrote: One of the concerns we discussed earlier is the additional requirement: 'subrepo' usage should be same/similar to 'single repo' usage. This clearly is not possible with (A). [we have to buy into the sub-repo model and use it appropriately].

i.e be aware which commands work with subrepos by default - and which ones need '-s' for subrepo

I think most users only need to pull and update which will do the right thing with subrepo.

- and then be aware that buildsystem you pull is not the latest [only the latest commited to petsc-dev] etc.

This is the right model anyway. If a BuildSystem patch is required by petsc-dev, there should be a commit in the petsc-dev repo stating that..

But this is not going to happen. If Matt fixes something in BuildSystem (for some reason not specific to something in petsc-dev he is working on) he will not doing it petsc-dev he will do it in BuildSystem. Basically you are asking Matt to do all his work in a very unnatural way because you chose to put BuildSystem into petsc-dev; this won't happen people instinctively violate unnatural ways of doing things. Barry