6.4 stable branch gone wild

classic Classic list List threaded Threaded
13 messages Options
Reply | Threaded
Open this post in threaded view
|

6.4 stable branch gone wild

tbonfort
Hi Devs,

There's been a mixup made when merging a pull request, which has
resulted in the master branch being merged into our stable 6.4 branch
(i.e. both our 6.4 and master branches are now quasi identical, which
is not ok).
I'm at loss as to how to revert this change cleanly without rewriting
the history of branch-6-4, can anyone help? It would seem that
75f548e77e1 was merged into 6.4 in a9eb0c5081 instead of being
cherrypicked, thus bringing in the whole master branch instead of only
the fix to mapfile.c.

regards,
thomas
_______________________________________________
mapserver-dev mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/mapserver-dev
Reply | Threaded
Open this post in threaded view
|

Re: 6.4 stable branch gone wild

Even Rouault
Thomas,

I'm far from being a git expert, but I can see 2 options:

* "git revert -m 1 a9eb0c5081" : this will create a revert commit for the
merge commit and should undo all unwanted changes. The drawback is that the
history of the branch will look messy.

* "git reset --hard d5f0aee829f9fef52a3d92a993baf2eac98e919a" to reset to the
last commit before the merge commit. But I don't think you can push that
directly into branch-6-4. So branch-6-4 should likely be deleted on the server
first before that being pushed (or git push --force ?). Drawback: this has the
potential to cause issues for people who have forked the mapserver git
repository when they'll pull/push next.

Some interesting info there :
https://htcondor-wiki.cs.wisc.edu/index.cgi/wiki?p=RevertingBadMerges

> Hi Devs,
>
> There's been a mixup made when merging a pull request, which has
> resulted in the master branch being merged into our stable 6.4 branch
> (i.e. both our 6.4 and master branches are now quasi identical, which
> is not ok).
> I'm at loss as to how to revert this change cleanly without rewriting
> the history of branch-6-4, can anyone help? It would seem that
> 75f548e77e1 was merged into 6.4 in a9eb0c5081 instead of being
> cherrypicked, thus bringing in the whole master branch instead of only
> the fix to mapfile.c.
>
> regards,
> thomas
> _______________________________________________
> mapserver-dev mailing list
> [hidden email]
> http://lists.osgeo.org/mailman/listinfo/mapserver-dev

--
Geospatial professional services
http://even.rouault.free.fr/services.html
_______________________________________________
mapserver-dev mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/mapserver-dev
Reply | Threaded
Open this post in threaded view
|

Re: 6.4 stable branch gone wild

Daniel Morissette
In reply to this post by tbonfort
This may be a bit off-topic and of no use to help resolve the current
situation, but I'm trying to understand and learn from what happened,
and  looking at the commits that you refer to on github it is not
obvious what happened, they look like simple two-liner merges:

https://github.com/mapserver/mapserver/commit/75f548e77e1

https://github.com/mapserver/mapserver/commit/a9eb0c5081

How can a github illiterate like me avoid making this kind of mistakes
by approving what looks like a simple pull request? What should we look
at? If this is a RTFMS then please point us (all github newbies) to the
light.

Thanks

Daniel


On 13-11-28 8:11 AM, thomas bonfort wrote:

> Hi Devs,
>
> There's been a mixup made when merging a pull request, which has
> resulted in the master branch being merged into our stable 6.4 branch
> (i.e. both our 6.4 and master branches are now quasi identical, which
> is not ok).
> I'm at loss as to how to revert this change cleanly without rewriting
> the history of branch-6-4, can anyone help? It would seem that
> 75f548e77e1 was merged into 6.4 in a9eb0c5081 instead of being
> cherrypicked, thus bringing in the whole master branch instead of only
> the fix to mapfile.c.
>
> regards,
> thomas
> _______________________________________________
> mapserver-dev mailing list
> [hidden email]
> http://lists.osgeo.org/mailman/listinfo/mapserver-dev
>


--
Daniel Morissette
http://www.mapgears.com/
Provider of Professional MapServer Support since 2000

_______________________________________________
mapserver-dev mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/mapserver-dev
Reply | Threaded
Open this post in threaded view
|

Re: 6.4 stable branch gone wild

Even Rouault
Le jeudi 28 novembre 2013 16:01:40, Daniel Morissette a écrit :

> This may be a bit off-topic and of no use to help resolve the current
> situation, but I'm trying to understand and learn from what happened,
> and  looking at the commits that you refer to on github it is not
> obvious what happened, they look like simple two-liner merges:
>
> https://github.com/mapserver/mapserver/commit/75f548e77e1
>
> https://github.com/mapserver/mapserver/commit/a9eb0c5081
>
> How can a github illiterate like me avoid making this kind of mistakes
> by approving what looks like a simple pull request? What should we look
> at? If this is a RTFMS then please point us (all github newbies) to the
> light.

Actually,  I had the same question and I don't think that a9eb0c5081 is the
merge that caused the issue. I now think this is due to
https://github.com/mapserver/mapserver/commit/b081e9f3b3d968b8c3b720a976e2a432b7295ed3

"git log --graph" shows that this commit results of a very long lasting branch
(presumably master) being merged into the current branch.



*   commit b081e9f3b3d968b8c3b720a976e2a432b7295ed3
|\  Merge: b7cb571 d4d3e22
| | Author: Paul Ramsey <[hidden email]>
| | Date:   Mon Nov 25 10:21:08 2013 -0800
| |
| |     Merge pull request #4803 from tbonfort/force2d
| |    
| |     PostGIS 2.1 ST_Force_2D is depreciated
| |  
| * commit d4d3e2205eeadd3f4b2ad60d803aa7c1836d8dee
| | Author: Thomas Bonfort <[hidden email]>
| | Date:   Tue Nov 5 10:04:49 2013 +0100
| |
| |     use version specific postgis force2d function (#4803)
| |    
| |     also adds a PROCESSING "FORCE2D=no" to avoid using force2d altogether
(#4024)
| |  
| * commit 3a6fa42289cad3596ccb2437167d7de5b0265f75
| | Author: Thomas Bonfort <[hidden email]>
| | Date:   Tue Nov 5 09:27:31 2013 +0100
| |
| |     use st_force2d instead of deprecated st_force_2d (#4803)
| |  
* | commit b7cb5712b6bcdb23b2aa8214207bfd2feff50465
| | Author: George Chang <[hidden email]>
| | Date:   Wed Nov 6 12:39:48 2013 -0800
| |
| |     Allows floating point NoData/NULLVALUEs for floating point images.
| |    
* |   commit 624940251f6dad286f6bc6033497527a298c02b9
|\ \  Merge: a9eb0c5 bdaf0cc
| | | Author: Thomas Bonfort <[hidden email]>
| | | Date:   Wed Nov 20 14:59:03 2013 +0100
| | |
| | |     Merge branch 'branch-6-4'
| | |    
| | |     Conflicts:
| | |           maputil.c
| | |           msautotest


The "gitk" gui also shows that.

But I'm surprised not to see an explicit "Merge master into branch-6-4" commit
somewhere...


>
> Thanks
>
> Daniel
>
> On 13-11-28 8:11 AM, thomas bonfort wrote:
> > Hi Devs,
> >
> > There's been a mixup made when merging a pull request, which has
> > resulted in the master branch being merged into our stable 6.4 branch
> > (i.e. both our 6.4 and master branches are now quasi identical, which
> > is not ok).
> > I'm at loss as to how to revert this change cleanly without rewriting
> > the history of branch-6-4, can anyone help? It would seem that
> > 75f548e77e1 was merged into 6.4 in a9eb0c5081 instead of being
> > cherrypicked, thus bringing in the whole master branch instead of only
> > the fix to mapfile.c.
> >
> > regards,
> > thomas
> > _______________________________________________
> > mapserver-dev mailing list
> > [hidden email]
> > http://lists.osgeo.org/mailman/listinfo/mapserver-dev

--
Geospatial professional services
http://even.rouault.free.fr/services.html
_______________________________________________
mapserver-dev mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/mapserver-dev
Reply | Threaded
Open this post in threaded view
|

Re: 6.4 stable branch gone wild

tbonfort
Even, a checkout at b7cb5712b6bcd which is right before that postgis
pull request, contains fontcache.c , i.e. already has master merged.
Daniel, I don't think this happened through the github interface, but
rather through the git command line (Steve, please confirm). My guess
is that there was a "git merge 75f548e77" instead of a "git
cherry-pick 75f548e77" run on the 6.4 branch (75f548e77 at that time
was master plus 1 commit)

On 28 November 2013 18:23, Even Rouault <[hidden email]> wrote:

> Le jeudi 28 novembre 2013 16:01:40, Daniel Morissette a écrit :
>> This may be a bit off-topic and of no use to help resolve the current
>> situation, but I'm trying to understand and learn from what happened,
>> and  looking at the commits that you refer to on github it is not
>> obvious what happened, they look like simple two-liner merges:
>>
>> https://github.com/mapserver/mapserver/commit/75f548e77e1
>>
>> https://github.com/mapserver/mapserver/commit/a9eb0c5081
>>
>> How can a github illiterate like me avoid making this kind of mistakes
>> by approving what looks like a simple pull request? What should we look
>> at? If this is a RTFMS then please point us (all github newbies) to the
>> light.
>
> Actually,  I had the same question and I don't think that a9eb0c5081 is the
> merge that caused the issue. I now think this is due to
> https://github.com/mapserver/mapserver/commit/b081e9f3b3d968b8c3b720a976e2a432b7295ed3
>
> "git log --graph" shows that this commit results of a very long lasting branch
> (presumably master) being merged into the current branch.
>
>
>
> *   commit b081e9f3b3d968b8c3b720a976e2a432b7295ed3
> |\  Merge: b7cb571 d4d3e22
> | | Author: Paul Ramsey <[hidden email]>
> | | Date:   Mon Nov 25 10:21:08 2013 -0800
> | |
> | |     Merge pull request #4803 from tbonfort/force2d
> | |
> | |     PostGIS 2.1 ST_Force_2D is depreciated
> | |
> | * commit d4d3e2205eeadd3f4b2ad60d803aa7c1836d8dee
> | | Author: Thomas Bonfort <[hidden email]>
> | | Date:   Tue Nov 5 10:04:49 2013 +0100
> | |
> | |     use version specific postgis force2d function (#4803)
> | |
> | |     also adds a PROCESSING "FORCE2D=no" to avoid using force2d altogether
> (#4024)
> | |
> | * commit 3a6fa42289cad3596ccb2437167d7de5b0265f75
> | | Author: Thomas Bonfort <[hidden email]>
> | | Date:   Tue Nov 5 09:27:31 2013 +0100
> | |
> | |     use st_force2d instead of deprecated st_force_2d (#4803)
> | |
> * | commit b7cb5712b6bcdb23b2aa8214207bfd2feff50465
> | | Author: George Chang <[hidden email]>
> | | Date:   Wed Nov 6 12:39:48 2013 -0800
> | |
> | |     Allows floating point NoData/NULLVALUEs for floating point images.
> | |
> * |   commit 624940251f6dad286f6bc6033497527a298c02b9
> |\ \  Merge: a9eb0c5 bdaf0cc
> | | | Author: Thomas Bonfort <[hidden email]>
> | | | Date:   Wed Nov 20 14:59:03 2013 +0100
> | | |
> | | |     Merge branch 'branch-6-4'
> | | |
> | | |     Conflicts:
> | | |           maputil.c
> | | |           msautotest
>
>
> The "gitk" gui also shows that.
>
> But I'm surprised not to see an explicit "Merge master into branch-6-4" commit
> somewhere...
>
>
>>
>> Thanks
>>
>> Daniel
>>
>> On 13-11-28 8:11 AM, thomas bonfort wrote:
>> > Hi Devs,
>> >
>> > There's been a mixup made when merging a pull request, which has
>> > resulted in the master branch being merged into our stable 6.4 branch
>> > (i.e. both our 6.4 and master branches are now quasi identical, which
>> > is not ok).
>> > I'm at loss as to how to revert this change cleanly without rewriting
>> > the history of branch-6-4, can anyone help? It would seem that
>> > 75f548e77e1 was merged into 6.4 in a9eb0c5081 instead of being
>> > cherrypicked, thus bringing in the whole master branch instead of only
>> > the fix to mapfile.c.
>> >
>> > regards,
>> > thomas
>> > _______________________________________________
>> > mapserver-dev mailing list
>> > [hidden email]
>> > http://lists.osgeo.org/mailman/listinfo/mapserver-dev
>
> --
> Geospatial professional services
> http://even.rouault.free.fr/services.html
> _______________________________________________
> mapserver-dev mailing list
> [hidden email]
> http://lists.osgeo.org/mailman/listinfo/mapserver-dev
_______________________________________________
mapserver-dev mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/mapserver-dev
Reply | Threaded
Open this post in threaded view
|

Re: 6.4 stable branch gone wild

Lime, Steve D (MNIT)
The msIO_printf pull request merge was done through the github interface, not the command line. Seemed simple enough.

Steve
________________________________________
From: [hidden email] [[hidden email]] on behalf of thomas bonfort [[hidden email]]
Sent: Thursday, November 28, 2013 11:40 AM
To: Even Rouault
Cc: MapServer Dev Mailing List; Daniel Morissette
Subject: Re: [mapserver-dev] 6.4 stable branch gone wild

Even, a checkout at b7cb5712b6bcd which is right before that postgis
pull request, contains fontcache.c , i.e. already has master merged.
Daniel, I don't think this happened through the github interface, but
rather through the git command line (Steve, please confirm). My guess
is that there was a "git merge 75f548e77" instead of a "git
cherry-pick 75f548e77" run on the 6.4 branch (75f548e77 at that time
was master plus 1 commit)

On 28 November 2013 18:23, Even Rouault <[hidden email]> wrote:

> Le jeudi 28 novembre 2013 16:01:40, Daniel Morissette a écrit :
>> This may be a bit off-topic and of no use to help resolve the current
>> situation, but I'm trying to understand and learn from what happened,
>> and  looking at the commits that you refer to on github it is not
>> obvious what happened, they look like simple two-liner merges:
>>
>> https://github.com/mapserver/mapserver/commit/75f548e77e1
>>
>> https://github.com/mapserver/mapserver/commit/a9eb0c5081
>>
>> How can a github illiterate like me avoid making this kind of mistakes
>> by approving what looks like a simple pull request? What should we look
>> at? If this is a RTFMS then please point us (all github newbies) to the
>> light.
>
> Actually,  I had the same question and I don't think that a9eb0c5081 is the
> merge that caused the issue. I now think this is due to
> https://github.com/mapserver/mapserver/commit/b081e9f3b3d968b8c3b720a976e2a432b7295ed3
>
> "git log --graph" shows that this commit results of a very long lasting branch
> (presumably master) being merged into the current branch.
>
>
>
> *   commit b081e9f3b3d968b8c3b720a976e2a432b7295ed3
> |\  Merge: b7cb571 d4d3e22
> | | Author: Paul Ramsey <[hidden email]>
> | | Date:   Mon Nov 25 10:21:08 2013 -0800
> | |
> | |     Merge pull request #4803 from tbonfort/force2d
> | |
> | |     PostGIS 2.1 ST_Force_2D is depreciated
> | |
> | * commit d4d3e2205eeadd3f4b2ad60d803aa7c1836d8dee
> | | Author: Thomas Bonfort <[hidden email]>
> | | Date:   Tue Nov 5 10:04:49 2013 +0100
> | |
> | |     use version specific postgis force2d function (#4803)
> | |
> | |     also adds a PROCESSING "FORCE2D=no" to avoid using force2d altogether
> (#4024)
> | |
> | * commit 3a6fa42289cad3596ccb2437167d7de5b0265f75
> | | Author: Thomas Bonfort <[hidden email]>
> | | Date:   Tue Nov 5 09:27:31 2013 +0100
> | |
> | |     use st_force2d instead of deprecated st_force_2d (#4803)
> | |
> * | commit b7cb5712b6bcdb23b2aa8214207bfd2feff50465
> | | Author: George Chang <[hidden email]>
> | | Date:   Wed Nov 6 12:39:48 2013 -0800
> | |
> | |     Allows floating point NoData/NULLVALUEs for floating point images.
> | |
> * |   commit 624940251f6dad286f6bc6033497527a298c02b9
> |\ \  Merge: a9eb0c5 bdaf0cc
> | | | Author: Thomas Bonfort <[hidden email]>
> | | | Date:   Wed Nov 20 14:59:03 2013 +0100
> | | |
> | | |     Merge branch 'branch-6-4'
> | | |
> | | |     Conflicts:
> | | |           maputil.c
> | | |           msautotest
>
>
> The "gitk" gui also shows that.
>
> But I'm surprised not to see an explicit "Merge master into branch-6-4" commit
> somewhere...
>
>
>>
>> Thanks
>>
>> Daniel
>>
>> On 13-11-28 8:11 AM, thomas bonfort wrote:
>> > Hi Devs,
>> >
>> > There's been a mixup made when merging a pull request, which has
>> > resulted in the master branch being merged into our stable 6.4 branch
>> > (i.e. both our 6.4 and master branches are now quasi identical, which
>> > is not ok).
>> > I'm at loss as to how to revert this change cleanly without rewriting
>> > the history of branch-6-4, can anyone help? It would seem that
>> > 75f548e77e1 was merged into 6.4 in a9eb0c5081 instead of being
>> > cherrypicked, thus bringing in the whole master branch instead of only
>> > the fix to mapfile.c.
>> >
>> > regards,
>> > thomas
>> > _______________________________________________
>> > mapserver-dev mailing list
>> > [hidden email]
>> > http://lists.osgeo.org/mailman/listinfo/mapserver-dev
>
> --
> Geospatial professional services
> http://even.rouault.free.fr/services.html
> _______________________________________________
> mapserver-dev mailing list
> [hidden email]
> http://lists.osgeo.org/mailman/listinfo/mapserver-dev
_______________________________________________
mapserver-dev mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/mapserver-dev

_______________________________________________
mapserver-dev mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/mapserver-dev
Reply | Threaded
Open this post in threaded view
|

Re: 6.4 stable branch gone wild

Lime, Steve D (MNIT)
Note that I didn't try to merge that pull request back into the 6.4 branch, at least not intentionally. As far as I know only master was updated.

Steve
________________________________________
From: [hidden email] [[hidden email]] on behalf of Lime, Steve D (MNIT) [[hidden email]]
Sent: Friday, November 29, 2013 8:35 AM
To: thomas bonfort; Even Rouault
Cc: MapServer Dev Mailing List; Daniel Morissette
Subject: Re: [mapserver-dev] 6.4 stable branch gone wild

The msIO_printf pull request merge was done through the github interface, not the command line. Seemed simple enough.

Steve
________________________________________
From: [hidden email] [[hidden email]] on behalf of thomas bonfort [[hidden email]]
Sent: Thursday, November 28, 2013 11:40 AM
To: Even Rouault
Cc: MapServer Dev Mailing List; Daniel Morissette
Subject: Re: [mapserver-dev] 6.4 stable branch gone wild

Even, a checkout at b7cb5712b6bcd which is right before that postgis
pull request, contains fontcache.c , i.e. already has master merged.
Daniel, I don't think this happened through the github interface, but
rather through the git command line (Steve, please confirm). My guess
is that there was a "git merge 75f548e77" instead of a "git
cherry-pick 75f548e77" run on the 6.4 branch (75f548e77 at that time
was master plus 1 commit)

On 28 November 2013 18:23, Even Rouault <[hidden email]> wrote:

> Le jeudi 28 novembre 2013 16:01:40, Daniel Morissette a écrit :
>> This may be a bit off-topic and of no use to help resolve the current
>> situation, but I'm trying to understand and learn from what happened,
>> and  looking at the commits that you refer to on github it is not
>> obvious what happened, they look like simple two-liner merges:
>>
>> https://github.com/mapserver/mapserver/commit/75f548e77e1
>>
>> https://github.com/mapserver/mapserver/commit/a9eb0c5081
>>
>> How can a github illiterate like me avoid making this kind of mistakes
>> by approving what looks like a simple pull request? What should we look
>> at? If this is a RTFMS then please point us (all github newbies) to the
>> light.
>
> Actually,  I had the same question and I don't think that a9eb0c5081 is the
> merge that caused the issue. I now think this is due to
> https://github.com/mapserver/mapserver/commit/b081e9f3b3d968b8c3b720a976e2a432b7295ed3
>
> "git log --graph" shows that this commit results of a very long lasting branch
> (presumably master) being merged into the current branch.
>
>
>
> *   commit b081e9f3b3d968b8c3b720a976e2a432b7295ed3
> |\  Merge: b7cb571 d4d3e22
> | | Author: Paul Ramsey <[hidden email]>
> | | Date:   Mon Nov 25 10:21:08 2013 -0800
> | |
> | |     Merge pull request #4803 from tbonfort/force2d
> | |
> | |     PostGIS 2.1 ST_Force_2D is depreciated
> | |
> | * commit d4d3e2205eeadd3f4b2ad60d803aa7c1836d8dee
> | | Author: Thomas Bonfort <[hidden email]>
> | | Date:   Tue Nov 5 10:04:49 2013 +0100
> | |
> | |     use version specific postgis force2d function (#4803)
> | |
> | |     also adds a PROCESSING "FORCE2D=no" to avoid using force2d altogether
> (#4024)
> | |
> | * commit 3a6fa42289cad3596ccb2437167d7de5b0265f75
> | | Author: Thomas Bonfort <[hidden email]>
> | | Date:   Tue Nov 5 09:27:31 2013 +0100
> | |
> | |     use st_force2d instead of deprecated st_force_2d (#4803)
> | |
> * | commit b7cb5712b6bcdb23b2aa8214207bfd2feff50465
> | | Author: George Chang <[hidden email]>
> | | Date:   Wed Nov 6 12:39:48 2013 -0800
> | |
> | |     Allows floating point NoData/NULLVALUEs for floating point images.
> | |
> * |   commit 624940251f6dad286f6bc6033497527a298c02b9
> |\ \  Merge: a9eb0c5 bdaf0cc
> | | | Author: Thomas Bonfort <[hidden email]>
> | | | Date:   Wed Nov 20 14:59:03 2013 +0100
> | | |
> | | |     Merge branch 'branch-6-4'
> | | |
> | | |     Conflicts:
> | | |           maputil.c
> | | |           msautotest
>
>
> The "gitk" gui also shows that.
>
> But I'm surprised not to see an explicit "Merge master into branch-6-4" commit
> somewhere...
>
>
>>
>> Thanks
>>
>> Daniel
>>
>> On 13-11-28 8:11 AM, thomas bonfort wrote:
>> > Hi Devs,
>> >
>> > There's been a mixup made when merging a pull request, which has
>> > resulted in the master branch being merged into our stable 6.4 branch
>> > (i.e. both our 6.4 and master branches are now quasi identical, which
>> > is not ok).
>> > I'm at loss as to how to revert this change cleanly without rewriting
>> > the history of branch-6-4, can anyone help? It would seem that
>> > 75f548e77e1 was merged into 6.4 in a9eb0c5081 instead of being
>> > cherrypicked, thus bringing in the whole master branch instead of only
>> > the fix to mapfile.c.
>> >
>> > regards,
>> > thomas
>> > _______________________________________________
>> > mapserver-dev mailing list
>> > [hidden email]
>> > http://lists.osgeo.org/mailman/listinfo/mapserver-dev
>
> --
> Geospatial professional services
> http://even.rouault.free.fr/services.html
> _______________________________________________
> mapserver-dev mailing list
> [hidden email]
> http://lists.osgeo.org/mailman/listinfo/mapserver-dev
_______________________________________________
mapserver-dev mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/mapserver-dev

_______________________________________________
mapserver-dev mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/mapserver-dev


_______________________________________________
mapserver-dev mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/mapserver-dev
Reply | Threaded
Open this post in threaded view
|

Re: 6.4 stable branch gone wild

Lime, Steve D (MNIT)
In reply to this post by Even Rouault
Seems like option 1 is the most straight forward IMHO... Wish I understood how this happened.

Steve

________________________________________
From: [hidden email] [[hidden email]] on behalf of Even Rouault [[hidden email]]
Sent: Thursday, November 28, 2013 7:52 AM
To: [hidden email]
Subject: Re: [mapserver-dev] 6.4 stable branch gone wild

Thomas,

I'm far from being a git expert, but I can see 2 options:

* "git revert -m 1 a9eb0c5081" : this will create a revert commit for the
merge commit and should undo all unwanted changes. The drawback is that the
history of the branch will look messy.

* "git reset --hard d5f0aee829f9fef52a3d92a993baf2eac98e919a" to reset to the
last commit before the merge commit. But I don't think you can push that
directly into branch-6-4. So branch-6-4 should likely be deleted on the server
first before that being pushed (or git push --force ?). Drawback: this has the
potential to cause issues for people who have forked the mapserver git
repository when they'll pull/push next.

Some interesting info there :
https://htcondor-wiki.cs.wisc.edu/index.cgi/wiki?p=RevertingBadMerges

> Hi Devs,
>
> There's been a mixup made when merging a pull request, which has
> resulted in the master branch being merged into our stable 6.4 branch
> (i.e. both our 6.4 and master branches are now quasi identical, which
> is not ok).
> I'm at loss as to how to revert this change cleanly without rewriting
> the history of branch-6-4, can anyone help? It would seem that
> 75f548e77e1 was merged into 6.4 in a9eb0c5081 instead of being
> cherrypicked, thus bringing in the whole master branch instead of only
> the fix to mapfile.c.
>
> regards,
> thomas
> _______________________________________________
> mapserver-dev mailing list
> [hidden email]
> http://lists.osgeo.org/mailman/listinfo/mapserver-dev

--
Geospatial professional services
http://even.rouault.free.fr/services.html
_______________________________________________
mapserver-dev mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/mapserver-dev


_______________________________________________
mapserver-dev mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/mapserver-dev
Reply | Threaded
Open this post in threaded view
|

Re: 6.4 stable branch gone wild

tbonfort
Well, I should have looked into my own mess rather than point fingers
erroneously, and would like to humbly apologize about that. After
another few hours of going through the commits and wondering how/what
happened, `git reflog` showed me that when I was updating my checkout
for applying https://github.com/mapserver/mapserver/pull/4805, I made
a typo and pulled origin/master instead of origin/branch-6-4 into my
local 6-4 branch, effectively merging master into 6.4. Given I had
just previously backported the 6.4 fixes into master (by merging
branch-6-4 into master), the erroneous pull effectively just
fast-forwarded branch-6-4 to be the same as master (i.e. I don't have
a commit I can revert).
While it is easy to revert back to the 6.4 branch as it should be, I
can't see how I can do that without rewriting the history, which would
affect people who have pulled the 6.4 branch since november 20th.
(From the network graph, this would mean Paul, Karl and myself (cc'd))
I'll leave this here for one day in case someone comes up with a
better solution, and if not will apply that change, and followup with
instructions for Paul, Karl and any other people who will have to
update their repository for the rewritten history.

Sorry folks,
thomas

On 30 November 2013 18:07, Lime, Steve D (MNIT) <[hidden email]> wrote:

> Seems like option 1 is the most straight forward IMHO... Wish I understood how this happened.
>
> Steve
>
> ________________________________________
> From: [hidden email] [[hidden email]] on behalf of Even Rouault [[hidden email]]
> Sent: Thursday, November 28, 2013 7:52 AM
> To: [hidden email]
> Subject: Re: [mapserver-dev] 6.4 stable branch gone wild
>
> Thomas,
>
> I'm far from being a git expert, but I can see 2 options:
>
> * "git revert -m 1 a9eb0c5081" : this will create a revert commit for the
> merge commit and should undo all unwanted changes. The drawback is that the
> history of the branch will look messy.
>
> * "git reset --hard d5f0aee829f9fef52a3d92a993baf2eac98e919a" to reset to the
> last commit before the merge commit. But I don't think you can push that
> directly into branch-6-4. So branch-6-4 should likely be deleted on the server
> first before that being pushed (or git push --force ?). Drawback: this has the
> potential to cause issues for people who have forked the mapserver git
> repository when they'll pull/push next.
>
> Some interesting info there :
> https://htcondor-wiki.cs.wisc.edu/index.cgi/wiki?p=RevertingBadMerges
>
>> Hi Devs,
>>
>> There's been a mixup made when merging a pull request, which has
>> resulted in the master branch being merged into our stable 6.4 branch
>> (i.e. both our 6.4 and master branches are now quasi identical, which
>> is not ok).
>> I'm at loss as to how to revert this change cleanly without rewriting
>> the history of branch-6-4, can anyone help? It would seem that
>> 75f548e77e1 was merged into 6.4 in a9eb0c5081 instead of being
>> cherrypicked, thus bringing in the whole master branch instead of only
>> the fix to mapfile.c.
>>
>> regards,
>> thomas
>> _______________________________________________
>> mapserver-dev mailing list
>> [hidden email]
>> http://lists.osgeo.org/mailman/listinfo/mapserver-dev
>
> --
> Geospatial professional services
> http://even.rouault.free.fr/services.html
> _______________________________________________
> mapserver-dev mailing list
> [hidden email]
> http://lists.osgeo.org/mailman/listinfo/mapserver-dev
>
>
> _______________________________________________
> mapserver-dev mailing list
> [hidden email]
> http://lists.osgeo.org/mailman/listinfo/mapserver-dev
_______________________________________________
mapserver-dev mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/mapserver-dev
Reply | Threaded
Open this post in threaded view
|

Re: 6.4 stable branch gone wild

Lime, Steve D (MNIT)
Hey Thomas: I'm glad you found the root cause, it's good to understand that. As always
your hard work on the project is very much appreciated so no worries at all!!!

Steve

________________________________________
From: [hidden email] [[hidden email]] on behalf of thomas bonfort [[hidden email]]
Sent: Monday, December 02, 2013 7:16 AM
To: [hidden email]
Cc: [hidden email]
Subject: Re: [mapserver-dev] 6.4 stable branch gone wild

Well, I should have looked into my own mess rather than point fingers
erroneously, and would like to humbly apologize about that. After
another few hours of going through the commits and wondering how/what
happened, `git reflog` showed me that when I was updating my checkout
for applying https://github.com/mapserver/mapserver/pull/4805, I made
a typo and pulled origin/master instead of origin/branch-6-4 into my
local 6-4 branch, effectively merging master into 6.4. Given I had
just previously backported the 6.4 fixes into master (by merging
branch-6-4 into master), the erroneous pull effectively just
fast-forwarded branch-6-4 to be the same as master (i.e. I don't have
a commit I can revert).
While it is easy to revert back to the 6.4 branch as it should be, I
can't see how I can do that without rewriting the history, which would
affect people who have pulled the 6.4 branch since november 20th.
(From the network graph, this would mean Paul, Karl and myself (cc'd))
I'll leave this here for one day in case someone comes up with a
better solution, and if not will apply that change, and followup with
instructions for Paul, Karl and any other people who will have to
update their repository for the rewritten history.

Sorry folks,
thomas

On 30 November 2013 18:07, Lime, Steve D (MNIT) <[hidden email]> wrote:

> Seems like option 1 is the most straight forward IMHO... Wish I understood how this happened.
>
> Steve
>
> ________________________________________
> From: [hidden email] [[hidden email]] on behalf of Even Rouault [[hidden email]]
> Sent: Thursday, November 28, 2013 7:52 AM
> To: [hidden email]
> Subject: Re: [mapserver-dev] 6.4 stable branch gone wild
>
> Thomas,
>
> I'm far from being a git expert, but I can see 2 options:
>
> * "git revert -m 1 a9eb0c5081" : this will create a revert commit for the
> merge commit and should undo all unwanted changes. The drawback is that the
> history of the branch will look messy.
>
> * "git reset --hard d5f0aee829f9fef52a3d92a993baf2eac98e919a" to reset to the
> last commit before the merge commit. But I don't think you can push that
> directly into branch-6-4. So branch-6-4 should likely be deleted on the server
> first before that being pushed (or git push --force ?). Drawback: this has the
> potential to cause issues for people who have forked the mapserver git
> repository when they'll pull/push next.
>
> Some interesting info there :
> https://htcondor-wiki.cs.wisc.edu/index.cgi/wiki?p=RevertingBadMerges
>
>> Hi Devs,
>>
>> There's been a mixup made when merging a pull request, which has
>> resulted in the master branch being merged into our stable 6.4 branch
>> (i.e. both our 6.4 and master branches are now quasi identical, which
>> is not ok).
>> I'm at loss as to how to revert this change cleanly without rewriting
>> the history of branch-6-4, can anyone help? It would seem that
>> 75f548e77e1 was merged into 6.4 in a9eb0c5081 instead of being
>> cherrypicked, thus bringing in the whole master branch instead of only
>> the fix to mapfile.c.
>>
>> regards,
>> thomas
>> _______________________________________________
>> mapserver-dev mailing list
>> [hidden email]
>> http://lists.osgeo.org/mailman/listinfo/mapserver-dev
>
> --
> Geospatial professional services
> http://even.rouault.free.fr/services.html
> _______________________________________________
> mapserver-dev mailing list
> [hidden email]
> http://lists.osgeo.org/mailman/listinfo/mapserver-dev
>
>
> _______________________________________________
> mapserver-dev mailing list
> [hidden email]
> http://lists.osgeo.org/mailman/listinfo/mapserver-dev
_______________________________________________
mapserver-dev mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/mapserver-dev


_______________________________________________
mapserver-dev mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/mapserver-dev
Reply | Threaded
Open this post in threaded view
|

Re: 6.4 stable branch gone wild

tbonfort
I've pushed the changes to the github repository... for anyone that
has pulled branch-6-4 since november 20th, the easiest is to delete
your working branch-6-4 and start it back from scratch:

#move off off the current branch-6-4
git checkout master
#delete current branch-6-4
git branch -d branch-6-4
#update our remote
git fetch origin #supposing origin is mapserver/mapserver
#checkout the updated 6-4 branch
git checkout --track origin/branch-6-4

catch me on irc if you need help

regards,
thomas

On 2 December 2013 15:06, Lime, Steve D (MNIT) <[hidden email]> wrote:

> Hey Thomas: I'm glad you found the root cause, it's good to understand that. As always
> your hard work on the project is very much appreciated so no worries at all!!!
>
> Steve
>
> ________________________________________
> From: [hidden email] [[hidden email]] on behalf of thomas bonfort [[hidden email]]
> Sent: Monday, December 02, 2013 7:16 AM
> To: [hidden email]
> Cc: [hidden email]
> Subject: Re: [mapserver-dev] 6.4 stable branch gone wild
>
> Well, I should have looked into my own mess rather than point fingers
> erroneously, and would like to humbly apologize about that. After
> another few hours of going through the commits and wondering how/what
> happened, `git reflog` showed me that when I was updating my checkout
> for applying https://github.com/mapserver/mapserver/pull/4805, I made
> a typo and pulled origin/master instead of origin/branch-6-4 into my
> local 6-4 branch, effectively merging master into 6.4. Given I had
> just previously backported the 6.4 fixes into master (by merging
> branch-6-4 into master), the erroneous pull effectively just
> fast-forwarded branch-6-4 to be the same as master (i.e. I don't have
> a commit I can revert).
> While it is easy to revert back to the 6.4 branch as it should be, I
> can't see how I can do that without rewriting the history, which would
> affect people who have pulled the 6.4 branch since november 20th.
> (From the network graph, this would mean Paul, Karl and myself (cc'd))
> I'll leave this here for one day in case someone comes up with a
> better solution, and if not will apply that change, and followup with
> instructions for Paul, Karl and any other people who will have to
> update their repository for the rewritten history.
>
> Sorry folks,
> thomas
>
> On 30 November 2013 18:07, Lime, Steve D (MNIT) <[hidden email]> wrote:
>> Seems like option 1 is the most straight forward IMHO... Wish I understood how this happened.
>>
>> Steve
>>
>> ________________________________________
>> From: [hidden email] [[hidden email]] on behalf of Even Rouault [[hidden email]]
>> Sent: Thursday, November 28, 2013 7:52 AM
>> To: [hidden email]
>> Subject: Re: [mapserver-dev] 6.4 stable branch gone wild
>>
>> Thomas,
>>
>> I'm far from being a git expert, but I can see 2 options:
>>
>> * "git revert -m 1 a9eb0c5081" : this will create a revert commit for the
>> merge commit and should undo all unwanted changes. The drawback is that the
>> history of the branch will look messy.
>>
>> * "git reset --hard d5f0aee829f9fef52a3d92a993baf2eac98e919a" to reset to the
>> last commit before the merge commit. But I don't think you can push that
>> directly into branch-6-4. So branch-6-4 should likely be deleted on the server
>> first before that being pushed (or git push --force ?). Drawback: this has the
>> potential to cause issues for people who have forked the mapserver git
>> repository when they'll pull/push next.
>>
>> Some interesting info there :
>> https://htcondor-wiki.cs.wisc.edu/index.cgi/wiki?p=RevertingBadMerges
>>
>>> Hi Devs,
>>>
>>> There's been a mixup made when merging a pull request, which has
>>> resulted in the master branch being merged into our stable 6.4 branch
>>> (i.e. both our 6.4 and master branches are now quasi identical, which
>>> is not ok).
>>> I'm at loss as to how to revert this change cleanly without rewriting
>>> the history of branch-6-4, can anyone help? It would seem that
>>> 75f548e77e1 was merged into 6.4 in a9eb0c5081 instead of being
>>> cherrypicked, thus bringing in the whole master branch instead of only
>>> the fix to mapfile.c.
>>>
>>> regards,
>>> thomas
>>> _______________________________________________
>>> mapserver-dev mailing list
>>> [hidden email]
>>> http://lists.osgeo.org/mailman/listinfo/mapserver-dev
>>
>> --
>> Geospatial professional services
>> http://even.rouault.free.fr/services.html
>> _______________________________________________
>> mapserver-dev mailing list
>> [hidden email]
>> http://lists.osgeo.org/mailman/listinfo/mapserver-dev
>>
>>
>> _______________________________________________
>> mapserver-dev mailing list
>> [hidden email]
>> http://lists.osgeo.org/mailman/listinfo/mapserver-dev
> _______________________________________________
> mapserver-dev mailing list
> [hidden email]
> http://lists.osgeo.org/mailman/listinfo/mapserver-dev
>
>
_______________________________________________
mapserver-dev mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/mapserver-dev
Reply | Threaded
Open this post in threaded view
|

Re: 6.4 stable branch gone wild

Even Rouault
Selon thomas bonfort <[hidden email]>:

> I've pushed the changes to the github repository... for anyone that
> has pulled branch-6-4 since november 20th, the easiest is to delete
> your working branch-6-4 and start it back from scratch:
>
> #move off off the current branch-6-4
> git checkout master
> #delete current branch-6-4
> git branch -d branch-6-4
> #update our remote
> git fetch origin #supposing origin is mapserver/mapserver
> #checkout the updated 6-4 branch
> git checkout --track origin/branch-6-4
>
> catch me on irc if you need help

Hi Thomas,

Would you mind explaining the procedure you followed to recreate the branch with
the appropriate commits in it (hoping that you didn't need to write down each
commit sha1 and cherry-pick it...) ? I imagine it could be helpful in case this
happens to someone else.

Thanks

Even
_______________________________________________
mapserver-dev mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/mapserver-dev
Reply | Threaded
Open this post in threaded view
|

Re: 6.4 stable branch gone wild

tbonfort
Even,
There was no magic involved, all was done manually (there was very
little to be done anyways):
- cherry-picked one commit
- merged one pull-request (4803)

regards,
thomas

On 3 December 2013 13:11, Even Rouault <[hidden email]> wrote:

> Selon thomas bonfort <[hidden email]>:
>
>> I've pushed the changes to the github repository... for anyone that
>> has pulled branch-6-4 since november 20th, the easiest is to delete
>> your working branch-6-4 and start it back from scratch:
>>
>> #move off off the current branch-6-4
>> git checkout master
>> #delete current branch-6-4
>> git branch -d branch-6-4
>> #update our remote
>> git fetch origin #supposing origin is mapserver/mapserver
>> #checkout the updated 6-4 branch
>> git checkout --track origin/branch-6-4
>>
>> catch me on irc if you need help
>
> Hi Thomas,
>
> Would you mind explaining the procedure you followed to recreate the branch with
> the appropriate commits in it (hoping that you didn't need to write down each
> commit sha1 and cherry-pick it...) ? I imagine it could be helpful in case this
> happens to someone else.
>
> Thanks
>
> Even
_______________________________________________
mapserver-dev mailing list
[hidden email]
http://lists.osgeo.org/mailman/listinfo/mapserver-dev