Zachary Amsden wrote:> Is there any particular protocol we should use to avoid stomping on
> each others toes when working on particular patches? It is quite easy
> to generate double nasty conflicts when working with patches of patches.
>
> I'm currently working through the patch set starting at the top (my
> patch tool was unhappy with 002-sync-bitops), working down my todo list.
>
> bobo@linux:~/paravirt> hg status
> M 002-sync-bitops.patch
> M 003-remove-ring0-assumptions.patch
It's a bit of a pain. Maybe we should treat the patches as
semi-immutable; when you want to make a change, add a new patch on top
of the existing one, and then when we decide to merge we can smush it
all together again. Unfortunately it doesn't work if the new patches
break subsequent patches in the series.
Well, for now I'm working on the paravirt and Xen (02*-*) parts of the
series. No doubt once you start doing VMI bindings you'll want to add a
VMI series (I suggest numbering them 04*-* perhaps), and make other
changes to the paravirt ops.
Maybe the easiest thing to do is push all the early part of the series
upstream, and then edit an actual tree rather than a patch series,
leaving the job of making a new series for later. I'm not very excited
by the prospect of teasing another postable patch series out of a tree
though; we'd get a nicer result by maintaining patches from the start.
J