Re: [Quilt-dev] merging with upstream and adding new files

[Jean Delvare]

[Jean Delvare]
> You should really dedicate the subsection to a typical in place codebase
> update (and we don't care how the update is done, manually or through
> CVS, it doesn't matter).

[Jerome Lacoste]
> It matters in the sense that with CVS, if you do it wrong, you have to
> start from scratch. Hence my confusing footnote.

I still don't see how different it is from a non CVS-based baseline
update. For example, here is how I handle my local stack of kernel
patches:

1* I start from the latest Linus' -rc tree.

2* I push patches as they are sent to me. After some times I have a
significant number of patches in my stack. Some of them eventually go to
Greg KH, then Andrew Morton, then Linus Torvalds.

3* When a newer -rc tree is available, I want to use it as my new
baseline. I pop all patches, then apply the incremental -rc patch to my
tree.

4* I delete all patches previously in my series which made it to Linus'
already, as they are now part of my baseline.

5* Finally I push all the remaining patches in the series, fixing
conflicts as needed.

This third step is very similar to a "cvs update" or equivalent, isn't
it? And if I'd do it the wrong way around (e.g. applying the baseline
patch before popping all the series patches), I'd screw everything up
just as a cvs update would do.

[Jerome Lacoste]
> Agree. But on the other side merging with upstream is an important
> part of the usage of quilt. Isn't it?

Agreed, and documenting it is very needed. But let's keep the
documentation concise, so that the reader actually understands how easy
it is. Updating the codebase is a simple, 4 step process: pop all
patches, update the codebase (whatever the method is), delete all
redundant patches, push all remaining patches. I don't think we have to
say more than just that.

Thanks,
--
Jean Delvare