I just pushed master into emacs-27

I just pushed master into emacs-27

[Nicolas Petton]

[-- Attachment #1: Type: text/plain, Size: 372 bytes --]

Hi,

I just realized that I pushed master into the emacs-27 branch... I must
have rebased against master out of habit, like I always do on some other
git repositories.

I'm really sorry about this, it has made a big mess in the emacs-27
branch.

I thought branches were protected against force-pushing, so I'm not sure
what is the best way for me to fix my mistake.

Nico

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]

Re: I just pushed master into emacs-27

[Amin Bandali]

[-- Attachment #1: Type: text/plain, Size: 1239 bytes --]

Hello,

Nicolas Petton <nico@petton.fr> writes:

> Hi,
>
> I just realized that I pushed master into the emacs-27 branch... I must
> have rebased against master out of habit, like I always do on some other
> git repositories.
>
> I'm really sorry about this, it has made a big mess in the emacs-27
> branch.
>
> I thought branches were protected against force-pushing, so I'm not sure
> what is the best way for me to fix my mistake.
>
> Nico

I suggest we all STOP pushing anything to any branches of the emacs.git
repository *immediately*, until Eli and our git experts see and chime in
this thread, and until we figure this out.

I think the most important question at this point is: can this change be
effectively and reasonably reversed, or is the emacs.git repository and
its emacs-27 branch on Savannah irreversibly messed up / cluttered?

If the answer to the above is no, we will probably have to get Savannah
hackers and the FSF sysadmins involved, to help us restore Savannah's
copy of emacs.git to an earlier snapshot if any.  Documentation on the
Savannah administration wiki [0] suggests that Savannah data is backed
up by the FSF daily.  I hope that's accurate/up-to-date.

[0]: https://savannah.gnu.org/maintenance/Backup/

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 857 bytes --]

Re: I just pushed master into emacs-27

[Nicolas Petton]

[-- Attachment #1: Type: text/plain, Size: 506 bytes --]

Amin Bandali <bandali@gnu.org> writes:

Hi Amin,

> If the answer to the above is no, we will probably have to get Savannah
> hackers and the FSF sysadmins involved, to help us restore Savannah's
> copy of emacs.git to an earlier snapshot if any.  Documentation on the
> Savannah administration wiki [0] suggests that Savannah data is backed
> up by the FSF daily.  I hope that's accurate/up-to-date.

I still have the commits on my local git repo, which I can push
somewhere else/on another branch.

Nico

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]

Re: I just pushed master into emacs-27

[Nicolas Petton]

[-- Attachment #1: Type: text/plain, Size: 410 bytes --]

Amin Bandali <bandali@gnu.org> writes:

> I think the most important question at this point is: can this change be
> effectively and reasonably reversed, or is the emacs.git repository and
> its emacs-27 branch on Savannah irreversibly messed up / cluttered?

The branch should be protected against non fast-forward pushes.  Also,
the faulty commits to emacs-27 might have already been pulled by people.

Nico

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]

Re: I just pushed master into emacs-27

[Eli Zaretskii]

> From: Nicolas Petton <nico@petton.fr>
> Cc: Emacs Devel <emacs-devel@gnu.org>, Eli Zaretskii <eliz@gnu.org>
> Date: Thu, 30 Jul 2020 01:07:55 +0200
> 
> > I think the most important question at this point is: can this change be
> > effectively and reasonably reversed, or is the emacs.git repository and
> > its emacs-27 branch on Savannah irreversibly messed up / cluttered?
> 
> The branch should be protected against non fast-forward pushes.

It is.

Re: I just pushed master into emacs-27

[Stefan Monnier]

>> > I think the most important question at this point is: can this change be
>> > effectively and reasonably reversed, or is the emacs.git repository and
>> > its emacs-27 branch on Savannah irreversibly messed up / cluttered?
>> The branch should be protected against non fast-forward pushes.
> It is.

Indeed it is protected, but this push *was* a fast-forward.


        Stefan

Re: I just pushed master into emacs-27

[dick.r.chiang]

AB> I suggest we all STOP pushing anything to any branches of the
emacs.git AB> repository *immediately*

This is unduly alarmist.  What happens in other branches is
independent of the emacs-27 branch.

AB> its emacs-27 branch on Savannah irreversibly messed up /
cluttered?

Minimally, M. Petton (and probably others) will have a reflog commit with
emacs27's last pristine state before the force push, which he could "git
checkout -B" and re-force-push.  Such is the distributed ethos of git, a
virtue Linus never fails to remind us of.

Re: I just pushed master into emacs-27

[Amin Bandali]

[-- Attachment #1: Type: text/plain, Size: 1564 bytes --]

dick.r.chiang@gmail.com writes:

> AB> I suggest we all STOP pushing anything to any branches of the
> emacs.git AB> repository *immediately*
>
> This is unduly alarmist.  What happens in other branches is
> independent of the emacs-27 branch.
>

I've been under the impression that in a Git repo, branches are
lightweight pointers to commits stored in the repo.  As such, my thought
process was that if Nico's commits have had a more global impact (on the
entire emacs.git repo) than just emacs-27, we may want to consider
having Savannah hackers restore an earlier snapshot of the entire repo
on the server side.  With that in mind, my suggestion to temporarily
suspend any pushes to the repo was so that we minimize changes that may
have to be re-pushed if it is decided that restoring emacs.git on the
server is necessary.

That said, I'm no git expert, and I certainly was not trying to sound
"alarmist".  If all of this can 'simply' be reversed by a force push and
without incurring any costs to emacs.git's internal state on Savannah,
and as if this never happened, then great.  I'm no fan of incurring
unnecessary work onto the Savannah hackers and/or the FSF sysadmins.

>
> AB> its emacs-27 branch on Savannah irreversibly messed up /
> cluttered? 
>
> Minimally, M. Petton (and probably others) will have a reflog commit with
> emacs27's last pristine state before the force push, which he could "git
> checkout -B" and re-force-push.  Such is the distributed ethos of git, a
> virtue Linus never fails to remind us of.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 853 bytes --]

Re: I just pushed master into emacs-27

[Lars Ingebrigtsen]

Nicolas Petton <nico@petton.fr> writes:

> I just realized that I pushed master into the emacs-27 branch... I must
> have rebased against master out of habit, like I always do on some other
> git repositories.
>
> I'm really sorry about this, it has made a big mess in the emacs-27
> branch.

Oops.  My sympathies.

But perhaps this isn't really that serious...  I mean, if it had been
master, then that would have been a different story, but this is just on
a branch, so the history there isn't really that big a deal, so you can
just revert the whole thing?

If the history is important, would it make sense, for instance, just to
make a emacs-27-2 branch starting from just before the merge and then
continue off of that?

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: I just pushed master into emacs-27

[황병희]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> Nicolas Petton <nico@petton.fr> writes:
>
>> I just realized that I pushed master into the emacs-27 branch... I must
>> have rebased against master out of habit, like I always do on some other
>> git repositories.
>>
>> I'm really sorry about this, it has made a big mess in the emacs-27
>> branch.
>
> Oops.  My sympathies.
>
> But perhaps this isn't really that serious...  I mean, if it had been
> master, then that would have been a different story, but this is just on
> a branch, so the history there isn't really that big a deal, so you can
> just revert the whole thing?
>
> If the history is important, would it make sense, for instance, just to
> make a emacs-27-2 branch starting from just before the merge and then
> continue off of that?

Indeed, i also did think same with Lars, for a while.

Sincerely, Gnus fan Byung-Hee

-- 
^고맙습니다 _救濟蒼生_ 감사합니다_^))//

Re: I just pushed master into emacs-27

[Stefan Kangas]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> But perhaps this isn't really that serious...

Indeed, git is as good at allowing you to shoot yourself in the foot
as it is at gracefully recovering from this type of situation.

Stefan M did exactly the right thing here, IMO.  There should be no
need for concern if you haven't already rebased your work on top of
the incorrect emacs-27 branch.  (Also on master it would have made the
most sense to do what Stefan M did, I think.)

Best regards,
Stefan Kangas

Re: I just pushed master into emacs-27

[Bob Proulx]

[-- Attachment #1: Type: text/plain, Size: 2012 bytes --]

Hello Nico,

Amin Bandali alerted me to the issues here.  I am one of the Savannah
Hackers and can help with some manual git adjustment on the repository
side of things.

Nicolas Petton wrote:
> I just realized that I pushed master into the emacs-27 branch... I must
> have rebased against master out of habit, like I always do on some other
> git repositories.
> 
> I'm really sorry about this, it has made a big mess in the emacs-27
> branch.
> 
> I thought branches were protected against force-pushing, so I'm not sure
> what is the best way for me to fix my mistake.

Branches are protected against non-fastforward pushes.  Which is a
good thing for preventing loss of data due to accidents involving
forced pushes.  And is a good safety net for times like this too since
we are assured that the repository hasn't lost anything.

This is up to you to decide but I suggest that we manually reset the
emacs-27 branch to the commit hash that it was before this problematic
push.  It's true that for any downstream sandbox that has already
pulled from the branch that they will also need to manually reset
their local sandbox branch too.  But the earlier this is done the more
limited in scope of the users for which will need to take this action.

This is easily done on the superuser side of the repository.  For
paranoia and safety sake I would make a branch of the current emacs-27
branch named emacs-27-oops-2020-07-29 or something for a bit of time
just as a safety backup just in case.  It can be truly deleted later.
Then simply reset the commit of the emacs-27 branch back to the
previous known good one one.  Then further pushes can happen just as
normal on top of it.

The only thing I would need to know is what is the git commit hash to
reset the emacs-27 branch back to in order to be before this push?

Let me know if that is what you would want to do.  Or if there is a
different action you would prefer be taken instead.  We are happy to
help!

Bob

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: I just pushed master into emacs-27

[Stefan Monnier]

> This is up to you to decide but I suggest that we manually reset the
> emacs-27 branch to the commit hash that it was before this problematic
> push.

I already did that.


        Stefan

Re: I just pushed master into emacs-27

[Nicolas Petton]

[-- Attachment #1: Type: text/plain, Size: 99 bytes --]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

> I already did that.

Thank you Stephan.

Nico

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]

Re: I just pushed master into emacs-27

[Andy Moreton]

On Wed 29 Jul 2020, Stefan Monnier wrote:

>> This is up to you to decide but I suggest that we manually reset the
>> emacs-27 branch to the commit hash that it was before this problematic
>> push.
>
> I already did that.

Yet you did not bother to send an announce message to the list so that
everyone who is not reading this thread realises what has happened.

Changing history is noteworthy - why have you not informed people ?

    AndyM

Re: I just pushed master into emacs-27

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: "Emacs Devel" <emacs-devel@gnu.org>,  Eli Zaretskii <eliz@gnu.org>
> Date: Wed, 29 Jul 2020 22:44:43 -0400
> 
> > This is up to you to decide but I suggest that we manually reset the
> > emacs-27 branch to the commit hash that it was before this problematic
> > push.
> 
> I already did that.

Thanks for a quick fix.

Re: I just pushed master into emacs-27

[Nicolas Petton]

[-- Attachment #1: Type: text/plain, Size: 318 bytes --]

Bob Proulx <bob@proulx.com> writes:

> Hello Nico,

Hi Bob,

> This is up to you to decide but I suggest that we manually reset the
> emacs-27 branch to the commit hash that it was before this problematic
> push.

Yes, that's what I wanted to do, but couldn't as it was not a non
fast-forward operation.

Thanks,
Nico

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]

Re: I just pushed master into emacs-27

[Eli Zaretskii]

> From: Nicolas Petton <nico@petton.fr>
> Date: Wed, 29 Jul 2020 23:35:46 +0200
> Cc: Eli Zaretskii <eliz@gnu.org>
> 
> I just realized that I pushed master into the emacs-27 branch...

Are you sure you pushed?  I've did a "git pull", and I don't see any
traces of such a push.  What am I missing?

If the problem is local to your clone, then you can "git reset" to fix
it locally.

Re: I just pushed master into emacs-27

[Nicolas Petton]

[-- Attachment #1: Type: text/plain, Size: 227 bytes --]

Eli Zaretskii <eliz@gnu.org> writes:

> Are you sure you pushed?  I've did a "git pull", and I don't see any
> traces of such a push.  What am I missing?

Yes, quite sure.  Stephan just said he manually reset the branch.

Nico

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]