A bit of quilt

[Amit Kumar]

On Thu, Feb 02, 2017 at 10:51:37AM +0100, Greg KH wrote:
> On Thu, Feb 02, 2017 at 09:08:15AM +0000, Amit Kumar wrote:
> > On Thu, Feb 02, 2017 at 08:10:56AM +0100, Greg KH wrote:
> > > On Thu, Feb 02, 2017 at 05:11:32AM +0000, Amit Kumar wrote:
> > > > On Wed, Feb 01, 2017 at 01:54:16PM +0000, Amit Kumar wrote:
> > > > > Hi,
> > > > > As I know quilt is used by maintainers. But kernel source code is
> > > > > maintained in git repo. So I want to know how git and quilt work
> > > > > together.
> > > > I have found a video in which Mr. GregKH has explained how he applies
> > > > patches to the stable tree. But this video is short and needs several view to
> > > > understand what is going on.
> > > 
> > > Do you have questions about it?
> > I use git-send-email and msmtp combination.
> > I also use mutt and msmtp combination.
> > Could you provide me quilt mail and git-send-email configuration?
> 
> You already have git-send-email working, what do you need changed there?
> 
> As for quilt mail, what did you try that did not work?
> 
> > > > > In mutt I have seen that a mail sent by Mr. GregKH has  quilt mail as user
> > > > > agent and git-send-email as x-mailer. It means he is using
> > > > > git-send-email as a backend for quilt mail.
> > > > > 
> > > > > Last but not least, I think if a developer starts using quilt to
> > > > > maintain his diferent versions of a patch, it will ease a maintainer
> > > > > job.
> > > > I'm in the process of making developers available upto minute code under
> > > > change.So duplicate patch problem can be solved.
> > > 
> > > What duplicate patch problem?
> > Sometimes when a developer sends his patch. He receives a reply this
> > patch has been already submitted by another developer.
> > 
> > I think the reason is that when a developer starts working on a patch,
> > he has not bleeding edge copy of maintainers tree. There is a long
> > review cycle of patch which is required for a large project as Linux.
> 
> Define "long" :)
> 
> Of course there will be conflicts, that's just the nature of working on
> a distributed project where no one can "lock" any portion of the tree.
> Just redo your patch and move on.  Nothing complex there.
> 
> > > > So I request kernel experts their words.
> > > 
> > > What question do you have?
> > So, I have a solution. As patches are collected on patchwork.kernel.org.
> > While patches are under review, it can be tracked by a bot and show lines
> > of code,on a web page, which will be affected on the basis of currently
> > submitted patch. So, developers don't touch those lines of code.
> 
> Nope.  That would prevent others from doing work, which is never a good
> idea.
> 
> How often have you really hit this issue?  As someone who reviews more
> patches than anyone else in the kernel, I see it happen only very
> infrequently (i.e. less than 1% of the time.)
Yes. This is  only a view for developers which shows which area of tree is
expected to change and under process, but we can't force anyone to not
touch that area of code.
> 
> > I also propose in-queue branch(patches in queue to be applied) for maintainers, 
> > which will help a maintainer to know which patches has been selected by
> > other maintainer. I think there will be less conflicts.
> 
> Where are the conflicts you see happening?  Again, is this really a big
> problem that you are trying to solve here?  I haven't heard any other
> maintainer complain about it, you do know about linux-next, right?
> 
I see whenever linux-next maintainer merge, he complains about manual
resolution of conflicts.
> > If you help me answering few questions as I develop this system, I will
> > be grateful to you.
> 
> Don't work to solve a non-existant problem :)
>
Ok, I consider your suggestion and first involve in kernel development
process heavily and then watch for problem. 

Last but not least, Thank you for your replies.

> thanks,
> 
> greg k-h