Re: [PATCH 14/24] pnfs_submit: use fsdata to pass lseg

[Boaz Harrosh <bharrosh@panasas.com>]

On 06/10/2010 03:48 PM, Benny Halevy wrote:
> On Jun. 10, 2010, 15:45 +0300, Benny Halevy <bhalevy@panasas.com> wrote:
>> On 06/10/2010 01:33 PM, Fred Isaman wrote:
>>>
>>> On Jun 9, 2010, at 8:08 AM, Fred Isaman wrote:
>>>
>>>> On Wed, Jun 9, 2010 at 6:38 AM, Benny Halevy <bhalevy@panasas.com> wrote:
>>>>> Fred, how does that patch interact with
>>>>> 285052f pnfs_post_submit: Restore "pnfs: pnfs_do_flush"
>>>>> and the latter patches that depend on it?
>>>>>
>>>>> Benny
>>>>>
>>>>
>>>> They will have to be modified.  I'll look at that today.
>>>>
>>>> Fred
>>>
>>> OK, this is a general git question.  How in the world do I send in these modifications?
>>>
>>> Basically, because of the way we have pnfs-submit in the middle of our tree, I have a branch that looks like:
>>>
>>> A->B->C->D
>>>
>>> I've inserted my new patch F between And B, which requires a rebase of the subsequent patches:
>>>
>>> A->F->B'->C'->D'
>>>
>>> But that rebase is non-trivial, in particular for patch C (a block-layout patch), and I want to communicate the modifications I made.
>>>
>>> The best I have been able to come up with is to do the minimal obvious rebase, just sufficient to remove all the conflict markers,
>>> then add a following modification patch, so I would have something like:
>>>
>>> A->F->B'->C'->C''->D'
>>>
>>> and I could send in C''.  But this seems less than ideal, especially when you consider I have ~10 patches which would require this handling.
>>>
>>
>> You can either send your a patchset based on C and I can
>> rebase parts of it onto A and B or just send the clean rebased
>> patches from B and C and we can review the diff (B vs. B' and
>> C vs. C')
> 
> Oh and I apologize for having set a moving target for you.
> I'm about to release a tree with merged patches from Alexandros,
> Andy, and Boaz (merge conflicts should be minimal though)
> 
>>
>> Benny
>>

Fred do you have a git tree on the linux-nfs.org/ or at Netapp that's open on the net?
If so you could just publish your cleanup tree, Benny will "git remote add" your branch
and can then cherry-pick / rebase very easily. It's how we did it a few times.
(We still need patches for review though)

Boaz