pgsql: Branch refs/heads/REL9_1_STABLE was removed

On 26 June 2011 00:02, <pgsql@postgresql.org> wrote:

Branch refs/heads/REL9_1_STABLE was removed.

What just happened here?!

--
Thom Brown
Twitter: @darkixion
IRC (freenode): dark_ixion
Registered Linux user: #516935

EnterpriseDB UK: http://www.enterprisedb.com
The Enterprise PostgreSQL Company

On 06/25/2011 04:02 PM, pgsql@postgresql.org wrote:

Branch refs/heads/REL9_1_STABLE was removed.

Umm, I was trying to follow the directions here:
http://wiki.postgresql.org/wiki/Committing_with_Git: Making a new
release branch

and it messed up my local repo such that
git push --dry-run

was giving an error. Googling the solution seemed to be:
git push origin :refs/heads/REL9_1_STABLE

I thought that would only affect my local repo, but apparently it did
not :-(

Help please...

Joe

--
Joe Conway
credativ LLC: http://www.credativ.us
Linux, PostgreSQL, and general Open Source
Training, Service, Consulting, & 24x7 Support

On 06/25/2011 07:07 PM, Joe Conway wrote:

On 06/25/2011 04:02 PM, pgsql@postgresql.org wrote:

Branch refs/heads/REL9_1_STABLE was removed.

Umm, I was trying to follow the directions here:
http://wiki.postgresql.org/wiki/Committing_with_Git: Making a new
release branch

and it messed up my local repo such that
git push --dry-run

was giving an error. Googling the solution seemed to be:
git push origin :refs/heads/REL9_1_STABLE

I thought that would only affect my local repo, but apparently it did
not :-(

Why would you be making a new release branch? I don't understand that bit.

FYI,

git push origin :somewhere

pushes nothing to the destination, i.e. empties it out. See git-push man
page, examples section.

cheers

andrew

Andrew Dunstan wrote:

On 06/25/2011 07:07 PM, Joe Conway wrote:

On 06/25/2011 04:02 PM, pgsql@postgresql.org wrote:

Branch refs/heads/REL9_1_STABLE was removed.

Umm, I was trying to follow the directions here:
http://wiki.postgresql.org/wiki/Committing_with_Git: Making a new
release branch

and it messed up my local repo such that
git push --dry-run

was giving an error. Googling the solution seemed to be:
git push origin :refs/heads/REL9_1_STABLE

I thought that would only affect my local repo, but apparently it did
not :-(

Why would you be making a new release branch? I don't understand that bit.

FYI,

git push origin :somewhere

pushes nothing to the destination, i.e. empties it out. See git-push man
page, examples section.

I talked to Joe on the phone. He was confused by the git commit wiki
section where we talked about creating a release branch:

http://wiki.postgresql.org/wiki/Committing_with_Git#Making_a_new_release_branch_on_origin

He thought it was for creating a local release branch. I have udpated
the title to be clearer:

Making a new release branch on origin

He did a git push origin REL9_1_STABLE and all is well --- I can pull
his 9.1 changes just fine.

--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ It's impossible for everything to be true. +

On 06/25/2011 04:44 PM, Andrew Dunstan wrote:

On 06/25/2011 07:07 PM, Joe Conway wrote:

On 06/25/2011 04:02 PM, pgsql@postgresql.org wrote:

Branch refs/heads/REL9_1_STABLE was removed.

Umm, I was trying to follow the directions here:
http://wiki.postgresql.org/wiki/Committing_with_Git: Making a new
release branch

and it messed up my local repo such that
git push --dry-run

was giving an error. Googling the solution seemed to be:
git push origin :refs/heads/REL9_1_STABLE

I thought that would only affect my local repo, but apparently it did
not :-(

Why would you be making a new release branch? I don't understand that bit.

I was misunderstanding the wiki page when trying to create my own local
9.1 branch. Bruce just helped me restore the origin 9.1 branch. I
*think* all is well now.

Joe

--
Joe Conway
credativ LLC: http://www.credativ.us
Linux, PostgreSQL, and general Open Source
Training, Service, Consulting, & 24x7 Support

On Sun, Jun 26, 2011 at 02:00, Joe Conway <mail@joeconway.com> wrote:

On 06/25/2011 04:44 PM, Andrew Dunstan wrote:

On 06/25/2011 07:07 PM, Joe Conway wrote:

On 06/25/2011 04:02 PM, pgsql@postgresql.org wrote:

Branch refs/heads/REL9_1_STABLE was removed.

Umm, I was trying to follow the directions here:
http://wiki.postgresql.org/wiki/Committing_with_Git: Making a new
release branch

and it messed up my local repo such that
   git push --dry-run

was giving an error. Googling the solution seemed to be:
   git push origin :refs/heads/REL9_1_STABLE

I thought that would only affect my local repo, but apparently it did
not :-(

Why would you be making a new release branch? I don't understand that bit.

I was misunderstanding the wiki page when trying to create my own local
9.1 branch. Bruce just helped me restore the origin 9.1 branch. I
*think* all is well now.

We discussed earlier to potentially block the creation, and removal,
of branches on the origin server, to prevent mistakes like this. It
has only happened once in almost a year, so it's probably not
necessary - but I wanted to raise the option anyway in case people
forgot about it.

The downside would be that in order to create or drop a branch *when
intended* a committer would need someone from the infrastructure team
to temporarily switch off the branch-blocking setting, and then back
on..

--
 Magnus Hagander
 Me: http://www.hagander.net/
 Work: http://www.redpill-linpro.com/

On 06/28/2011 01:49 AM, Magnus Hagander wrote:

On Sun, Jun 26, 2011 at 02:00, Joe Conway<mail@joeconway.com> wrote:

On 06/25/2011 04:44 PM, Andrew Dunstan wrote:

On 06/25/2011 07:07 PM, Joe Conway wrote:

On 06/25/2011 04:02 PM, pgsql@postgresql.org wrote:

Branch refs/heads/REL9_1_STABLE was removed.

Umm, I was trying to follow the directions here:
http://wiki.postgresql.org/wiki/Committing_with_Git: Making a new
release branch

and it messed up my local repo such that
git push --dry-run

was giving an error. Googling the solution seemed to be:
git push origin :refs/heads/REL9_1_STABLE

I thought that would only affect my local repo, but apparently it did
not :-(

Why would you be making a new release branch? I don't understand that bit.

I was misunderstanding the wiki page when trying to create my own local
9.1 branch. Bruce just helped me restore the origin 9.1 branch. I
*think* all is well now.

We discussed earlier to potentially block the creation, and removal,
of branches on the origin server, to prevent mistakes like this. It
has only happened once in almost a year, so it's probably not
necessary - but I wanted to raise the option anyway in case people
forgot about it.

The downside would be that in order to create or drop a branch *when
intended* a committer would need someone from the infrastructure team
to temporarily switch off the branch-blocking setting, and then back
on..

I think it's probably a good idea, at least in the case of removal.
After all, how often will we intentionally drop a branch?

Incidentally, the trouble with what Joe did to recover is that he didn't
push exactly what he deleted, so the mail record doesn't contain his
commit on the 9.1 branch. Ideally he should have reverted his local
branch, pushed that, then recommitted his patch and repushed the branch.

cheers

andrew

Magnus Hagander <magnus@hagander.net> writes:

We discussed earlier to potentially block the creation, and removal,
of branches on the origin server, to prevent mistakes like this. It
has only happened once in almost a year, so it's probably not
necessary - but I wanted to raise the option anyway in case people
forgot about it.

The downside would be that in order to create or drop a branch *when
intended* a committer would need someone from the infrastructure team
to temporarily switch off the branch-blocking setting, and then back
on..

I think it would be sensible to block branch removal, as there's
basically never a scenario where we'd do that during current usage.
I'm not excited about blocking branch addition, although I worry
sooner or later somebody will accidentally push a private development
branch :-(. Is it possible to block only removal and not addition?

regards, tom lane

On Tue, Jun 28, 2011 at 14:56, Andrew Dunstan <andrew@dunslane.net> wrote:

On 06/28/2011 01:49 AM, Magnus Hagander wrote:

On Sun, Jun 26, 2011 at 02:00, Joe Conway<mail@joeconway.com>  wrote:

On 06/25/2011 04:44 PM, Andrew Dunstan wrote:

On 06/25/2011 07:07 PM, Joe Conway wrote:

On 06/25/2011 04:02 PM, pgsql@postgresql.org wrote:

Branch refs/heads/REL9_1_STABLE was removed.

Umm, I was trying to follow the directions here:
http://wiki.postgresql.org/wiki/Committing_with_Git: Making a new
release branch

and it messed up my local repo such that
   git push --dry-run

was giving an error. Googling the solution seemed to be:
   git push origin :refs/heads/REL9_1_STABLE

I thought that would only affect my local repo, but apparently it did
not :-(

Why would you be making a new release branch? I don't understand that
bit.

I was misunderstanding the wiki page when trying to create my own local
9.1 branch. Bruce just helped me restore the origin 9.1 branch. I
*think* all is well now.

We discussed earlier to potentially block the creation, and removal,
of branches on the origin server, to prevent mistakes like this. It
has only happened once in almost a year, so it's probably not
necessary - but I wanted to raise the option anyway in case people
forgot about it.

The downside would be that in order to create or drop a branch *when
intended* a committer would need someone from the infrastructure team
to temporarily switch off the branch-blocking setting, and then back
on..

I think it's probably a good idea, at least in the case of removal. After
all, how often will we intentionally drop a branch?

yeha. OTOH, how often do we intenrionally *create* a branch? About
once / year...

Incidentally, the trouble with what Joe did to recover is that he didn't
push exactly what he deleted, so the mail record doesn't contain his commit
on the 9.1 branch. Ideally he should have reverted his local branch, pushed
that, then recommitted his patch and repushed the branch.

Right. The idea behind such a feature would be to protect against
*mistakes*, not malice..

--
 Magnus Hagander
 Me: http://www.hagander.net/
 Work: http://www.redpill-linpro.com/

On Tue, Jun 28, 2011 at 16:39, Tom Lane <tgl@sss.pgh.pa.us> wrote:

Magnus Hagander <magnus@hagander.net> writes:

We discussed earlier to potentially block the creation, and removal,
of branches on the origin server, to prevent mistakes like this. It
has only happened once in almost a year, so it's probably not
necessary - but I wanted to raise the option anyway in case people
forgot about it.

The downside would be that in order to create or drop a branch *when
intended* a committer would need someone from the infrastructure team
to temporarily switch off the branch-blocking setting, and then back
on..

I think it would be sensible to block branch removal, as there's
basically never a scenario where we'd do that during current usage.
I'm not excited about blocking branch addition, although I worry
sooner or later somebody will accidentally push a private development
branch :-(.  Is it possible to block only removal and not addition?

Yes, I think so. Either way it'll require a small addition to the
scripts we're using, so I'll try to just turn it into two different
settings. I don't see offhand any reason why this shouldn't work.

--
 Magnus Hagander
 Me: http://www.hagander.net/
 Work: http://www.redpill-linpro.com/

On 06/28/2011 07:39 AM, Magnus Hagander wrote:

Incidentally, the trouble with what Joe did to recover is that he didn't
push exactly what he deleted, so the mail record doesn't contain his commit
on the 9.1 branch. Ideally he should have reverted his local branch, pushed
that, then recommitted his patch and repushed the branch.

Right. The idea behind such a feature would be to protect against
*mistakes*, not malice..

That *was* a mistake on my part, not malice.

In any case, I was shocked that I was able to do what I did, so I would
support something that prevents mistakes -- at least big ones such as
creating or dropping branches unintentionally. Part of the problem here
is that the people who know exactly how to recover are the same ones who
are not as likely to make mistakes, and vice-versa.

Joe

--
Joe Conway
credativ LLC: http://www.credativ.us
Linux, PostgreSQL, and general Open Source
Training, Service, Consulting, & 24x7 Support

On 06/28/2011 10:39 AM, Tom Lane wrote:

Magnus Hagander<magnus@hagander.net> writes:

We discussed earlier to potentially block the creation, and removal,
of branches on the origin server, to prevent mistakes like this. It
has only happened once in almost a year, so it's probably not
necessary - but I wanted to raise the option anyway in case people
forgot about it.
The downside would be that in order to create or drop a branch *when
intended* a committer would need someone from the infrastructure team
to temporarily switch off the branch-blocking setting, and then back
on..

I think it would be sensible to block branch removal, as there's
basically never a scenario where we'd do that during current usage.
I'm not excited about blocking branch addition, although I worry
sooner or later somebody will accidentally push a private development
branch :-(. Is it possible to block only removal and not addition?

+1. Spurious branch addition shouldn't cause us much pain - we'd just
remove the new branch. Unwanted deletion is more disruptive.

cheers

andrew

On Tue, Jun 28, 2011 at 16:52, Joe Conway <mail@joeconway.com> wrote:

On 06/28/2011 07:39 AM, Magnus Hagander wrote:

Incidentally, the trouble with what Joe did to recover is that he didn't
push exactly what he deleted, so the mail record doesn't contain his commit
on the 9.1 branch. Ideally he should have reverted his local branch, pushed
that, then recommitted his patch and repushed the branch.

Right. The idea behind such a feature would be to protect against
*mistakes*, not malice..

That *was* a mistake on my part, not malice.

Yes, I'm pretty sure nobody thinks anything else!

In any case, I was shocked that I was able to do what I did, so I would
support something that prevents mistakes -- at least big ones such as
creating or dropping branches unintentionally. Part of the problem here
is that the people who know exactly how to recover are the same ones who
are not as likely to make mistakes, and vice-versa.

Yeah.

Ok, I have the script updated and it was easy to block both creation
and removal - it's a simple on/off parameter to the script. We just
need consensus on if we want to block just removal, or both removal
and creation.

--
 Magnus Hagander
 Me: http://www.hagander.net/
 Work: http://www.redpill-linpro.com/

On Tue, Jun 28, 2011 at 11:05 AM, Andrew Dunstan <andrew@dunslane.net> wrote:

+1. Spurious branch addition shouldn't cause us much pain - we'd just remove
the new branch. Unwanted deletion is more disruptive.

How about if we allow addition only of branches matching
/^REL_[0-9_]+_STABLE$/ and disallow deletion of all branches? That
seems like it'd allow the one operation we will likely want to do with
any regularity (creating a new release branch once a year) without
going through hoops, while disallowing most of the problem cases.

The problem with allowing people to create branches and not remove
them is that someone might push a private branch and not be able to
get rid of it. But if we only allow creation of branches that look
like the branches that are supposed to be there, then that shouldn't
be a danger.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company

On 06/28/2011 08:05 AM, Magnus Hagander wrote:

On Tue, Jun 28, 2011 at 16:52, Joe Conway <mail@joeconway.com> wrote:

That *was* a mistake on my part, not malice.

Yes, I'm pretty sure nobody thinks anything else!

I would hope so, but the way you quoted me it was a little unclear ;-)

Ok, I have the script updated and it was easy to block both creation
and removal - it's a simple on/off parameter to the script. We just
need consensus on if we want to block just removal, or both removal
and creation.

As stated elsewhere on this thread, we normally only create official
branches at most once per year or so. Unless we think that is going to
change in this brave new world of git, I would vote to block both, but
make sure the process to unblock is well documented on the wiki.

Joe

--
Joe Conway
credativ LLC: http://www.credativ.us
Linux, PostgreSQL, and general Open Source
Training, Service, Consulting, & 24x7 Support

Robert Haas <robertmhaas@gmail.com> writes:

How about if we allow addition only of branches matching
/^REL_[0-9_]+_STABLE$/ and disallow deletion of all branches?

+1, if feasible.

regards, tom lane

On Jun 28, 2011 6:29 PM, "Tom Lane" <tgl@sss.pgh.pa.us> wrote:

Robert Haas <robertmhaas@gmail.com> writes:

How about if we allow addition only of branches matching
/^REL_[0-9_]+_STABLE$/ and disallow deletion of all branches?

+1, if feasible.

Pretty sure that's just a Small Matter Of Programming. I'll give it a try.

/Magnus

Excerpts from Tom Lane's message of mar jun 28 10:39:22 -0400 2011:

I think it would be sensible to block branch removal, as there's
basically never a scenario where we'd do that during current usage.
I'm not excited about blocking branch addition, although I worry
sooner or later somebody will accidentally push a private development
branch :-(. Is it possible to block only removal and not addition?

If we can tweak the thing, how about we only allow creating branches
that match a certain pattern, say ^REL_\d+_\d+_STABLE$?

--
Álvaro Herrera <alvherre@commandprompt.com>
The PostgreSQL Company - Command Prompt, Inc.
PostgreSQL Replication, Consulting, Custom Development, 24x7 support

On Tue, Jun 28, 2011 at 19:45, Alvaro Herrera
<alvherre@commandprompt.com> wrote:

Excerpts from Tom Lane's message of mar jun 28 10:39:22 -0400 2011:

I think it would be sensible to block branch removal, as there's
basically never a scenario where we'd do that during current usage.
I'm not excited about blocking branch addition, although I worry
sooner or later somebody will accidentally push a private development
branch :-(.  Is it possible to block only removal and not addition?

If we can tweak the thing, how about we only allow creating branches
that match a certain pattern, say ^REL_\d+_\d+_STABLE$?

I've put this in place - except I used ^REL\d+... and not what you
suggested, since that's how we name our branches :P

I'm going to push an actual valid branch, let's say 9.7, and then
remove it again, just to make sure things worked (with it installed I
cannot push an invalid branch, so i can't test the branch removal
block). So don't panic if you see that one :)

--
 Magnus Hagander
 Me: http://www.hagander.net/
 Work: http://www.redpill-linpro.com/

On Wed, Jun 29, 2011 at 16:30, Magnus Hagander <magnus@hagander.net> wrote:

On Tue, Jun 28, 2011 at 19:45, Alvaro Herrera
<alvherre@commandprompt.com> wrote:

Excerpts from Tom Lane's message of mar jun 28 10:39:22 -0400 2011:

I think it would be sensible to block branch removal, as there's
basically never a scenario where we'd do that during current usage.
I'm not excited about blocking branch addition, although I worry
sooner or later somebody will accidentally push a private development
branch :-(.  Is it possible to block only removal and not addition?

If we can tweak the thing, how about we only allow creating branches
that match a certain pattern, say ^REL_\d+_\d+_STABLE$?

I've put this in place - except I used ^REL\d+... and not what you
suggested, since that's how we name our branches :P

I'm going to push an actual valid branch, let's say 9.7, and then
remove it again, just to make sure things worked (with it installed I
cannot push an invalid branch, so i can't test the branch removal
block). So don't panic if you see that one :)

Ok, I'm done experimenting and this is now in production.

--
 Magnus Hagander
 Me: http://www.hagander.net/
 Work: http://www.redpill-linpro.com/

Excerpts from Magnus Hagander's message of mié jun 29 10:30:51 -0400 2011:

On Tue, Jun 28, 2011 at 19:45, Alvaro Herrera
<alvherre@commandprompt.com> wrote:

Excerpts from Tom Lane's message of mar jun 28 10:39:22 -0400 2011:

I think it would be sensible to block branch removal, as there's
basically never a scenario where we'd do that during current usage.
I'm not excited about blocking branch addition, although I worry
sooner or later somebody will accidentally push a private development
branch :-(.  Is it possible to block only removal and not addition?

If we can tweak the thing, how about we only allow creating branches
that match a certain pattern, say ^REL_\d+_\d+_STABLE$?

I've put this in place - except I used ^REL\d+... and not what you
suggested, since that's how we name our branches :P

Hah! It took me a while to notice the difference :-D

Thanks! I know I feel safer with this in place :-P

--
Álvaro Herrera <alvherre@commandprompt.com>
The PostgreSQL Company - Command Prompt, Inc.
PostgreSQL Replication, Consulting, Custom Development, 24x7 support

On Thu, Jun 30, 2011 at 12:03 AM, Alvaro Herrera <alvherre@commandprompt.com

wrote:

Excerpts from Magnus Hagander's message of mié jun 29 10:30:51 -0400 2011:

On Tue, Jun 28, 2011 at 19:45, Alvaro Herrera
<alvherre@commandprompt.com> wrote:

Excerpts from Tom Lane's message of mar jun 28 10:39:22 -0400 2011:

I think it would be sensible to block branch removal, as there's
basically never a scenario where we'd do that during current usage.
I'm not excited about blocking branch addition, although I worry
sooner or later somebody will accidentally push a private development
branch :-(. Is it possible to block only removal and not addition?

If we can tweak the thing, how about we only allow creating branches
that match a certain pattern, say ^REL_\d+_\d+_STABLE$?

I've put this in place - except I used ^REL\d+... and not what you
suggested, since that's how we name our branches :P

Thanks, +10.
This is cool and will avoid for sure future problems.

--
Michael Paquier
http://michael.otacoo.com