On Fri, Sep 23, 2011 at 09:21:20PM +0200, Jiri Slaby wrote:

On 09/23/2011 09:08 PM, Greg KH wrote:

quoted

On Fri, Sep 23, 2011 at 08:52:16PM +0200, Jiri Slaby wrote:

quoted

On 09/23/2011 12:46 AM, Greg KH wrote:

quoted

On Wed, Aug 31, 2011 at 09:24:56PM +0200, Jiri Slaby wrote:

quoted

Commit "TTY: serial, move locking in uart_close" moved the lock, but
omitted to update branches which unlock the lock. Now they try to
unlock the lock without holding it.

Signed-off-by: Jiri Slaby <redacted>
---
If possible, please, merge this into the patch mentioned above (it's
not upstream yet).

I can't do that,

Hmm, but what is the reason for that? I mean, why do you prefer a kernel
with broken history with respect to bisection? Per definition -next
doesn't mind rebases in subtrees. Or is this already in tty-linus branch
(I cannot check now, obviously)?

Because it is in my tree and I can't rebase it as others depend on it
(linux-next and others.)

linux-next doesn't mind if you rebase. That's exactly what it is for. To
test commits collected from #for-next branches and alter them if needed.
It merges whatever is in the current branch no matter what was there
some days ago.

Users who base their work on mine care if I rebase.

And so do I, it's just one of those rules, "Greg will not rebase his
trees" that makes for better development.

And yes, sometimes it does cause minor problems like this, but overall,
it's much easier for everyone involved.

But if there are more trees depending on the tree, then OK, I will live
with that ;).

For the tty tree, I really doubt it, but I am not sure (rumor has it
that some people are basing on it, but that might just be rumor.)  For
my staging and USB trees, I can't rebase as I know I have users for
those trees.

greg k-h