Confused - bisecting wireless-testing

Confused - bisecting wireless-testing

[Luis R. Rodriguez]

04:23  * mcgrof is confused, I just did a git bisect -i on
wireless-testing on some patch on Fri Jul 2 00:09:49 2010  and ended
up with a 2.6.34 top level
          Makefile
04:23 < mcgrof> it was git rebase -i ba17bc5e55ba541d2a8765fca53b6883b667ab21
04:23 < mcgrof> eh
04:24 < mcgrof> how am I supposed to bisect wl now
04:24 < mcgrof> the odd thing though is that the top commit is ancient
04:24 < mcgrof> http://paste.pocoo.org/show/242077/
04:24 < mcgrof> but the other ones are OK

I realize we should use wireless-2.6.git to bisect stable but I need
to bisect against recent patches that spans different master- tags
from you, I figured git bisecting wireless-testing would work now that
you are using a different method to move your tree forward but am I
wrong? Should I only bisect between master tags still?

  Luis

Re: Confused - bisecting wireless-testing

[Luis R. Rodriguez]

On Mon, Jul 26, 2010 at 8:27 PM, Luis R. Rodriguez <mcgrof@gmail.com> wrote:
> 04:23  * mcgrof is confused, I just did a git bisect -i on
> wireless-testing on some patch on Fri Jul 2 00:09:49 2010  and ended
> up with a 2.6.34 top level
>          Makefile
> 04:23 < mcgrof> it was git rebase -i ba17bc5e55ba541d2a8765fca53b6883b667ab21
> 04:23 < mcgrof> eh
> 04:24 < mcgrof> how am I supposed to bisect wl now
> 04:24 < mcgrof> the odd thing though is that the top commit is ancient
> 04:24 < mcgrof> http://paste.pocoo.org/show/242077/
> 04:24 < mcgrof> but the other ones are OK
>
> I realize we should use wireless-2.6.git to bisect stable but I need
> to bisect against recent patches that spans different master- tags
> from you, I figured git bisecting wireless-testing would work now that
> you are using a different method to move your tree forward but am I
> wrong? Should I only bisect between master tags still?

Heh I just realize what I said didn't make sense, what I meant was I
was *going* to bisect but in order to first find my first good
starting point I am going through the tree trying to find a good
starting point to bisect and to do that I use interactive rebase (git
rebase -i sha1sum) to specific sha1sums until I find a good compile
that works. Only after I've done this would I then bisect.

The interactive bisect on ba17bc5e55ba541d2a8765fca53b6883b667ab21
throws me back to 2.6.34 though, for some odd reason.

  Luis

Re: Confused - bisecting wireless-testing

[John W. Linville]

On Mon, Jul 26, 2010 at 08:27:06PM -0700, Luis R. Rodriguez wrote:
> 04:23  * mcgrof is confused, I just did a git bisect -i on
> wireless-testing on some patch on Fri Jul 2 00:09:49 2010  and ended
> up with a 2.6.34 top level
>           Makefile
> 04:23 < mcgrof> it was git rebase -i ba17bc5e55ba541d2a8765fca53b6883b667ab21
> 04:23 < mcgrof> eh
> 04:24 < mcgrof> how am I supposed to bisect wl now
> 04:24 < mcgrof> the odd thing though is that the top commit is ancient
> 04:24 < mcgrof> http://paste.pocoo.org/show/242077/
> 04:24 < mcgrof> but the other ones are OK
> 
> I realize we should use wireless-2.6.git to bisect stable but I need
> to bisect against recent patches that spans different master- tags
> from you, I figured git bisecting wireless-testing would work now that
> you are using a different method to move your tree forward but am I
> wrong? Should I only bisect between master tags still?

wireless-testing is a nasty mess when it comes to bisection.
I am considering a one-time rebase of wireless-testing after the
next -rc1, now that the current process is working reasonably well.
wireless-testing will continue to have a messy history, but it should
be a bit less nasty after a rebase.

As for right now, it sounds like you should just use wireless-next-2.6
instead?

John
-- 
John W. Linville		Someday the world will need a hero, and you
linville@tuxdriver.com			might be all we have.  Be ready.

Re: Confused - bisecting wireless-testing

[Luis R. Rodriguez]

On Tue, Jul 27, 2010 at 7:01 AM, John W. Linville
<linville@tuxdriver.com> wrote:
> On Mon, Jul 26, 2010 at 08:27:06PM -0700, Luis R. Rodriguez wrote:
>> 04:23  * mcgrof is confused, I just did a git bisect -i on
>> wireless-testing on some patch on Fri Jul 2 00:09:49 2010  and ended
>> up with a 2.6.34 top level
>>           Makefile
>> 04:23 < mcgrof> it was git rebase -i ba17bc5e55ba541d2a8765fca53b6883b667ab21
>> 04:23 < mcgrof> eh
>> 04:24 < mcgrof> how am I supposed to bisect wl now
>> 04:24 < mcgrof> the odd thing though is that the top commit is ancient
>> 04:24 < mcgrof> http://paste.pocoo.org/show/242077/
>> 04:24 < mcgrof> but the other ones are OK
>>
>> I realize we should use wireless-2.6.git to bisect stable but I need
>> to bisect against recent patches that spans different master- tags
>> from you, I figured git bisecting wireless-testing would work now that
>> you are using a different method to move your tree forward but am I
>> wrong? Should I only bisect between master tags still?
>
> wireless-testing is a nasty mess when it comes to bisection.
> I am considering a one-time rebase of wireless-testing after the
> next -rc1, now that the current process is working reasonably well.
> wireless-testing will continue to have a messy history, but it should
> be a bit less nasty after a rebase.
>
> As for right now, it sounds like you should just use wireless-next-2.6
> instead?

Oh OK, didn't realize that was bisectable, thanks. Also was unsure if
I'd get other subsystems's -next components. Is wireless-next.git just
the -next bits for 802.11 or does it also suck up net-next?

  Luis

Re: Confused - bisecting wireless-testing

[John W. Linville]

On Tue, Jul 27, 2010 at 08:56:54AM -0700, Luis R. Rodriguez wrote:
> On Tue, Jul 27, 2010 at 7:01 AM, John W. Linville
> <linville@tuxdriver.com> wrote:
> > On Mon, Jul 26, 2010 at 08:27:06PM -0700, Luis R. Rodriguez wrote:
> >> 04:23  * mcgrof is confused, I just did a git bisect -i on
> >> wireless-testing on some patch on Fri Jul 2 00:09:49 2010  and ended
> >> up with a 2.6.34 top level
> >>           Makefile
> >> 04:23 < mcgrof> it was git rebase -i ba17bc5e55ba541d2a8765fca53b6883b667ab21
> >> 04:23 < mcgrof> eh
> >> 04:24 < mcgrof> how am I supposed to bisect wl now
> >> 04:24 < mcgrof> the odd thing though is that the top commit is ancient
> >> 04:24 < mcgrof> http://paste.pocoo.org/show/242077/
> >> 04:24 < mcgrof> but the other ones are OK
> >>
> >> I realize we should use wireless-2.6.git to bisect stable but I need
> >> to bisect against recent patches that spans different master- tags
> >> from you, I figured git bisecting wireless-testing would work now that
> >> you are using a different method to move your tree forward but am I
> >> wrong? Should I only bisect between master tags still?
> >
> > wireless-testing is a nasty mess when it comes to bisection.
> > I am considering a one-time rebase of wireless-testing after the
> > next -rc1, now that the current process is working reasonably well.
> > wireless-testing will continue to have a messy history, but it should
> > be a bit less nasty after a rebase.
> >
> > As for right now, it sounds like you should just use wireless-next-2.6
> > instead?
> 
> Oh OK, didn't realize that was bisectable, thanks. Also was unsure if
> I'd get other subsystems's -next components. Is wireless-next.git just
> the -next bits for 802.11 or does it also suck up net-next?

Usually just the wireless bits, but I did just pull the bluetooth stuff
(which included some net-next stuff).

John
-- 
John W. Linville		Someday the world will need a hero, and you
linville@tuxdriver.com			might be all we have.  Be ready.

Re: Confused - bisecting wireless-testing

[Luis R. Rodriguez]

On Tue, Jul 27, 2010 at 9:04 AM, John W. Linville
<linville@tuxdriver.com> wrote:
> On Tue, Jul 27, 2010 at 08:56:54AM -0700, Luis R. Rodriguez wrote:
>> On Tue, Jul 27, 2010 at 7:01 AM, John W. Linville
>> <linville@tuxdriver.com> wrote:
>> > On Mon, Jul 26, 2010 at 08:27:06PM -0700, Luis R. Rodriguez wrote:
>> >> 04:23  * mcgrof is confused, I just did a git bisect -i on
>> >> wireless-testing on some patch on Fri Jul 2 00:09:49 2010  and ended
>> >> up with a 2.6.34 top level
>> >>           Makefile
>> >> 04:23 < mcgrof> it was git rebase -i ba17bc5e55ba541d2a8765fca53b6883b667ab21
>> >> 04:23 < mcgrof> eh
>> >> 04:24 < mcgrof> how am I supposed to bisect wl now
>> >> 04:24 < mcgrof> the odd thing though is that the top commit is ancient
>> >> 04:24 < mcgrof> http://paste.pocoo.org/show/242077/
>> >> 04:24 < mcgrof> but the other ones are OK
>> >>
>> >> I realize we should use wireless-2.6.git to bisect stable but I need
>> >> to bisect against recent patches that spans different master- tags
>> >> from you, I figured git bisecting wireless-testing would work now that
>> >> you are using a different method to move your tree forward but am I
>> >> wrong? Should I only bisect between master tags still?
>> >
>> > wireless-testing is a nasty mess when it comes to bisection.
>> > I am considering a one-time rebase of wireless-testing after the
>> > next -rc1, now that the current process is working reasonably well.
>> > wireless-testing will continue to have a messy history, but it should
>> > be a bit less nasty after a rebase.
>> >
>> > As for right now, it sounds like you should just use wireless-next-2.6
>> > instead?
>>
>> Oh OK, didn't realize that was bisectable, thanks. Also was unsure if
>> I'd get other subsystems's -next components. Is wireless-next.git just
>> the -next bits for 802.11 or does it also suck up net-next?
>
> Usually just the wireless bits, but I did just pull the bluetooth stuff
> (which included some net-next stuff).

Oh neat. OK so which net-next.git release was wireless-next.git based
on and what net-next.git release was bluetooth-next.git based on? Did
it bump your next-next.git or do you routinely pull next-next.git into
your own wireless-next.git?

  Luis

Re: Confused - bisecting wireless-testing

[John W. Linville]

On Tue, Jul 27, 2010 at 09:21:40AM -0700, Luis R. Rodriguez wrote:
> On Tue, Jul 27, 2010 at 9:04 AM, John W. Linville
> <linville@tuxdriver.com> wrote:
> > On Tue, Jul 27, 2010 at 08:56:54AM -0700, Luis R. Rodriguez wrote:
> >> On Tue, Jul 27, 2010 at 7:01 AM, John W. Linville
> >> <linville@tuxdriver.com> wrote:

> >> > As for right now, it sounds like you should just use wireless-next-2.6
> >> > instead?
> >>
> >> Oh OK, didn't realize that was bisectable, thanks. Also was unsure if
> >> I'd get other subsystems's -next components. Is wireless-next.git just
> >> the -next bits for 802.11 or does it also suck up net-next?
> >
> > Usually just the wireless bits, but I did just pull the bluetooth stuff
> > (which included some net-next stuff).
> 
> Oh neat. OK so which net-next.git release was wireless-next.git based
> on and what net-next.git release was bluetooth-next.git based on? Did
> it bump your next-next.git or do you routinely pull next-next.git into
> your own wireless-next.git?

I don't have those exact commit IDs.  How badly do you need them?

I don't normally pull net-next into wireless-next, but this release
not doing that has been a source of pain due to the __packed patch
in net-next and the addition of __attribute__ ((packed)) in various
places in wireless-next.  Since Marcel had already based on a net-next
that had that patch and since it is close to the end of the release
I decided it would be acceptable to pull it this time.

I have been considering making regular net-next pulls a standard
practice.  That would make my life easier, but would add some
non-wireless instability into wireless-testing (which pulls from
wireless-next).  I don't think that would be a big problem _most_
of the time, but I'm sure there would be some occasional pain from
doing that.

John
-- 
John W. Linville		Someday the world will need a hero, and you
linville@tuxdriver.com			might be all we have.  Be ready.

Re: Confused - bisecting wireless-testing

[Luis R. Rodriguez]

On Tue, Jul 27, 2010 at 9:31 AM, John W. Linville
<linville@tuxdriver.com> wrote:
> On Tue, Jul 27, 2010 at 09:21:40AM -0700, Luis R. Rodriguez wrote:
>> On Tue, Jul 27, 2010 at 9:04 AM, John W. Linville
>> <linville@tuxdriver.com> wrote:
>> > On Tue, Jul 27, 2010 at 08:56:54AM -0700, Luis R. Rodriguez wrote:
>> >> On Tue, Jul 27, 2010 at 7:01 AM, John W. Linville
>> >> <linville@tuxdriver.com> wrote:
>
>> >> > As for right now, it sounds like you should just use wireless-next-2.6
>> >> > instead?
>> >>
>> >> Oh OK, didn't realize that was bisectable, thanks. Also was unsure if
>> >> I'd get other subsystems's -next components. Is wireless-next.git just
>> >> the -next bits for 802.11 or does it also suck up net-next?
>> >
>> > Usually just the wireless bits, but I did just pull the bluetooth stuff
>> > (which included some net-next stuff).
>>
>> Oh neat. OK so which net-next.git release was wireless-next.git based
>> on and what net-next.git release was bluetooth-next.git based on? Did
>> it bump your next-next.git or do you routinely pull next-next.git into
>> your own wireless-next.git?
>
> I don't have those exact commit IDs.  How badly do you need them?

I don't at all, was just interested to know if they were indeed
different and how each differ in terms of pulling new net-next.git
bits from future next-(date -I) releases.

> I don't normally pull net-next into wireless-next, but this release
> not doing that has been a source of pain due to the __packed patch
> in net-next and the addition of __attribute__ ((packed)) in various
> places in wireless-next.  Since Marcel had already based on a net-next
> that had that patch and since it is close to the end of the release
> I decided it would be acceptable to pull it this time.

Ah I see.

> I have been considering making regular net-next pulls a standard
> practice.  That would make my life easier, but would add some
> non-wireless instability into wireless-testing (which pulls from
> wireless-next).  I don't think that would be a big problem _most_
> of the time, but I'm sure there would be some occasional pain from
> doing that.

Thanks for the details. Yeah indeed my concern over bisecting
wireless-next.git would have been testing other bleeding edge
non-802.11 / BT subsystem stuff, things in net core, but relying on
the master-(date -I) tags sufficed for me to bisect an issue down to
early June. Think I'll stick to that for now but I suppose we should
not fear using next-next.git as much given that we now have both
802.11 and BT

So I guess the next question is where does David base his net-next.git
off from? Will it only include 802.11 / Ethernet / Wimax / netcore /
etc next bits or is there more? I ask to know what I'm in for in case
I want to bisect from it or derivatives (wireless-next.git).

  Luis

Re: Confused - bisecting wireless-testing

[John W. Linville]

On Tue, Jul 27, 2010 at 09:54:20AM -0700, Luis R. Rodriguez wrote:

> So I guess the next question is where does David base his net-next.git
> off from? Will it only include 802.11 / Ethernet / Wimax / netcore /
> etc next bits or is there more? I ask to know what I'm in for in case
> I want to bisect from it or derivatives (wireless-next.git).

Generally he branches from linux-2.6 at around -rc1 (and sometimes
before) and then doesn't pull again until after the release.

-- 
John W. Linville		Someday the world will need a hero, and you
linville@tuxdriver.com			might be all we have.  Be ready.

Re: Confused - bisecting wireless-testing

[Luis R. Rodriguez]

On Tue, Jul 27, 2010 at 10:18 AM, John W. Linville
<linville@tuxdriver.com> wrote:
> On Tue, Jul 27, 2010 at 09:54:20AM -0700, Luis R. Rodriguez wrote:
>
>> So I guess the next question is where does David base his net-next.git
>> off from? Will it only include 802.11 / Ethernet / Wimax / netcore /
>> etc next bits or is there more? I ask to know what I'm in for in case
>> I want to bisect from it or derivatives (wireless-next.git).
>
> Generally he branches from linux-2.6 at around -rc1 (and sometimes
> before) and then doesn't pull again until after the release.

Thanks this helps a lot!

  Luis