Fwd: Re: [GIT PULL] PCI changes for v5.12

[Greg KH]

On Fri, Feb 26, 2021 at 06:02:07PM +0800, wuzhouhui wrote:
> On 2/26/21 3:07 PM, Greg KH wrote:
> > On Fri, Feb 26, 2021 at 12:54:16PM +0800, wuzhouhui wrote:
> > > I'm confused about what the Linus said. It seemed that the branch cannot
> > > be rebased after merged into linux-next under normal circumstances, why?
> > You should never rebase a public-facing branch.
> > 
> > > Linus found this issue by check commit time of commits, but I also found so
> > > many commit in mainline are too young to be merged into linux-next before
> > > merge to mainline. If the rebase is not recommended, why people like to
> > > do rebase after commits are merged into linux-next?
> > What commits have you found, and are you sure those are not just merged
> > through email submissions to Linus instead of a git tree pull?  Please
> > note that some subsystems still use email, like Andrew Morton's.
> Following commit log are copied from mainline
> 
> 14 hours    Merge tag 'pwm/for-5.12-rc1' of
> git://git.kernel.org/pub/scm/linux/kernel/git...HEADmaster    Linus
> Torvalds    7    -374/+65
> 14 hours    Merge tag 'for_linus' of
> git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost    Linus Torvalds   
> 22    -507/+1492
> 14 hours    Merge tag 'mips_5.12_1' of
> git://git.kernel.org/pub/scm/linux/kernel/git/mips...    Linus Torvalds   
> 8    -5/+190
> 14 hours    Merge tag 'drm-next-2021-02-26' of
> git://anongit.freedesktop.org/drm/drm    Linus Torvalds    68 -423/+1223
> 14 hours    Merge tag 'net-5.12-rc1' of
> git://git.kernel.org/pub/scm/linux/kernel/git/net...    Linus Torvalds   
> 68    -371/+734
> 14 hours    Merge tag 'acpi-5.12-rc1-3' of
> git://git.kernel.org/pub/scm/linux/kernel/git/...    Linus Torvalds    8   
> -27/+334


Take a look at these and think about why they couldn't be in linux-next
beforehand :)

You might want to look at non-merge commits...

> 16 hours    net: phy: micrel: set soft_reset callback to genphy_soft_reset
> for KSZ8081    Christian Melki    1    -0/+1
> 16 hours    Merge tag 'nds32-for-linux-5.12' of
> git://git.kernel.org/pub/scm/linux/kernel...    Linus Torvalds 4    -50/+5
> 16 hours    net: psample: Fix netlink skb length with tunnel info Chris
> Mi    1    -2/+2
> 16 hours    net: broadcom: bcm4908_enet: fix NAPI poll returned value   
> Rafał Miłecki    1    -0/+2
> 16 hours    net: broadcom: bcm4908_enet: fix RX path possible mem leak   
> Rafał Miłecki    1    -0/+1
> 
> We can see that, last non-merge commit is just 2 hours older than
> merge-commits committed
> by Linus, so these non-merge commits shouldn't contained in linux-next
> before merge into
> mainline. Does me  misunderstand something?

For bugfixes, yes, they can be accepted now for "high volume" subsystems
like networking.  You don't want those to be delayed, especially when
they are found to be fixes in the previous pull request for that
subsystem, right?

But what we do not want is a bunch of "quick, let's rebase our patches
on Linus's latest tree!" for normal commits as then we have no idea if
they really were tested out in linux-next or not.

thanks,

greg k-h