[U-Boot] Patch submission process question

[U-Boot] Patch submission process question

[Joe Hershberger]

Hi u-boot,

What is the recommended way to handle a patch series that depends on
another patch series (already sent but not integrated)?

How about a patch that was sent, that at the time didn't depend on
another series, but after review and changes, now depends on those
changes?

Even a bit more generally, some of the series are getting quite long
due to small dependencies.  This tends to put lots of unrelated
content into the same series based on small dependencies.  How is this
supposed to be dealt with?  Is it best to make long series or just to
send part of it and wait for it to be integrated before sending the
rest?

Thanks,
-Joe

[U-Boot] Patch submission process question

[Simon Schwarz]

Hi Joe,

On 08/16/2011 03:20 AM, Joe Hershberger wrote:
> Hi u-boot,
>
> What is the recommended way to handle a patch series that depends on
> another patch series (already sent but not integrated)?
>
AFAIK the recommend way is to note in the patch that it depends on 
another - best with some kind of link from gmane or a message-ID.

> How about a patch that was sent, that at the time didn't depend on
> another series, but after review and changes, now depends on those
> changes?
Then you have to note it in the version history that it now depends on 
another patch - don't forget to also note it in the normal merge message.

>
> Even a bit more generally, some of the series are getting quite long
> due to small dependencies.  This tends to put lots of unrelated
> content into the same series based on small dependencies.  How is this
> supposed to be dealt with?  Is it best to make long series or just to
> send part of it and wait for it to be integrated before sending the
> rest?
Hm, here U'am not really sure - I think I would send a complete series - 
but this is best answered by a custodian.

>
> Thanks,
> -Joe

Regards
Simon

[U-Boot] Patch submission process question

[Scott Wood]

On 08/16/2011 02:06 AM, Simon Schwarz wrote:
> Hi Joe,
> 
> On 08/16/2011 03:20 AM, Joe Hershberger wrote:
>> Hi u-boot,
>>
>> What is the recommended way to handle a patch series that depends on
>> another patch series (already sent but not integrated)?
>>
> AFAIK the recommend way is to note in the patch that it depends on 
> another - best with some kind of link from gmane or a message-ID.

A patchwork link would be ideal.

>> How about a patch that was sent, that at the time didn't depend on
>> another series, but after review and changes, now depends on those
>> changes?
> Then you have to note it in the version history that it now depends on 
> another patch - don't forget to also note it in the normal merge message.

Merge message?  If you mean the commit message (above the "---"), why?
It should not be committed before its dependencies.

>> Even a bit more generally, some of the series are getting quite long
>> due to small dependencies.  This tends to put lots of unrelated
>> content into the same series based on small dependencies.  How is this
>> supposed to be dealt with?  Is it best to make long series or just to
>> send part of it and wait for it to be integrated before sending the
>> rest?
> Hm, here U'am not really sure - I think I would send a complete series - 
> but this is best answered by a custodian.

If it's getting very long (the threshold is obviously subjective), I'd
do the "send part of it and wait" approach.  Especially if there are a
lot of respins.

-Scott

[U-Boot] Patch submission process question

[Joe Hershberger]

On Tue, Aug 16, 2011 at 11:56 AM, Scott Wood <scottwood@freescale.com> wrote:
> On 08/16/2011 02:06 AM, Simon Schwarz wrote:
>> On 08/16/2011 03:20 AM, Joe Hershberger wrote:
>>> Even a bit more generally, some of the series are getting quite long
>>> due to small dependencies.  This tends to put lots of unrelated
>>> content into the same series based on small dependencies.  How is this
>>> supposed to be dealt with?  Is it best to make long series or just to
>>> send part of it and wait for it to be integrated before sending the
>>> rest?
>> Hm, here U'am not really sure - I think I would send a complete series -
>> but this is best answered by a custodian.
>
> If it's getting very long (the threshold is obviously subjective), I'd
> do the "send part of it and wait" approach.  Especially if there are a
> lot of respins.

How about if a series largely doesn't change but has one or two
patches change or maybe one or two patches added to the end?  Is it
preferable to resend all of the patches in the series marked as a new
version or to only send the changed / added patches?

Thanks,
-Joe

(Sorry about the double send, Scott).

[U-Boot] Patch submission process question

[Simon Schwarz]

On 08/16/2011 06:56 PM, Scott Wood wrote:
<snip>
>>> How about a patch that was sent, that at the time didn't depend on
>>> another series, but after review and changes, now depends on those
>>> changes?
>> Then you have to note it in the version history that it now depends on
>> another patch - don't forget to also note it in the normal merge message.
>
> Merge message?  If you mean the commit message (above the "---"), why?
> It should not be committed before its dependencies.
Sorry meant commit message. True - so just note it in the version 
history and - for convenience - in the cover letter?

Regards
Simon