threads / discuss / 30017

Has anyone written a hook to block fast forward merges to a branch?

Subject: Has anyone written a hook to block fast forward merges to a branch?

## tl;dr

6 messages between Mar 21, 2012 and Mar 21, 2012.

replies: 5people: 3as markdown or json

Chris Patti· Mar 21, 2012, 18:22 UTC · lore

I know there are hooks out there to block various other kinds of change, but I was wondering if anyone had specifically ever written one to block fast forward merges.

Thanks, -Chris

-- 
Christopher Patti - Geek At Large | GTalk: cpatti@gmail.com | AIM:
chrisfeohpatti | P: (260) 54PATTI
"Technology challenges art, art inspires technology." - John Lasseter, Pixar
Jeff King· Mar 21, 2012, 18:40 UTC · re: Chris Patti · lore

Re: Has anyone written a hook to block fast forward merges to a branch?

On Wed, Mar 21, 2012 at 02:22:46PM -0400, Chris Patti wrote:
> I know there are hooks out there to block various other kinds of
> change, but I was wondering if anyone had specifically ever written
> one to block fast forward merges.

How would you detect them? If I do a fast-forward merge, the result is indistinguishable from simply building the commits directly on the branch in the first place.

If you mean "nobody should ever build commits directly on master, but always be merging in commits from topic branches", then that is easier to enforce; just look for non-merge commits down the --first-parent line, and reject if you find any. I think you could do it with:

  git rev-list --max-parents=1 --first-parent $OLD..$NEW

If that produces output, then somebody build directly on master (or did a backwards master->topic merge, but if you care about first-parents lineage, then that is equally bad).

-Peff
Neal Kreitzinger· Mar 21, 2012, 20:32 UTC · re: Chris Patti · lore

Re: Has anyone written a hook to block fast forward merges to a branch?

On 3/21/2012 1:22 PM, Chris Patti wrote:
> I know there are hooks out there to block various other kinds of
> change, but I was wondering if anyone had specifically ever written
> one to block fast forward merges.
>
Maybe git-merge --no-ff would also be helpful to "do what you want".

(I'm not sure why you would tell people they can't do their work really fast or rebase to make it look like they did, but it seems that's what you are also implying.)

v/r, neal

Jeff King· Mar 21, 2012, 20:43 UTC · re: Neal Kreitzinger · lore

Re: Has anyone written a hook to block fast forward merges to a branch?

On Wed, Mar 21, 2012 at 03:32:10PM -0500, Neal Kreitzinger wrote:
Show 6 quoted lines
> On 3/21/2012 1:22 PM, Chris Patti wrote:
> >I know there are hooks out there to block various other kinds of
> >change, but I was wondering if anyone had specifically ever written
> >one to block fast forward merges.
> >
> Maybe git-merge --no-ff would also be helpful to "do what you want".

When I wrote my response, I assumed the intent was that people _should_ be using --no-ff, but that he wanted to enforce it via hook during a push to a central repository.

Reading it again, I'm not sure whether that is the case or not. If not, the disregard my original response. :)

-Peff
Neal Kreitzinger· Mar 21, 2012, 21:14 UTC · re: Jeff King · lore

Re: Has anyone written a hook to block fast forward merges to a branch?

On 3/21/2012 3:43 PM, Jeff King wrote:
Show 17 quoted lines
> On Wed, Mar 21, 2012 at 03:32:10PM -0500, Neal Kreitzinger wrote:
>
>> On 3/21/2012 1:22 PM, Chris Patti wrote:
>>> I know there are hooks out there to block various other kinds of
>>> change, but I was wondering if anyone had specifically ever
>>> written one to block fast forward merges.
>>>
>> Maybe git-merge --no-ff would also be helpful to "do what you
>> want".
>
> When I wrote my response, I assumed the intent was that people
> _should_ be using --no-ff, but that he wanted to enforce it via hook
> during a push to a central repository.
>
> Reading it again, I'm not sure whether that is the case or not. If
> not, the disregard my original response. :)
>

Would your script only work at "push time", or would it also work at "commit time"? I'm not sure at what point he thinks he wants to enforce it. I don't think git yet knows the info your script is asking for at "commit time" (ie, pre-commit hook and its brethren), and by the time it does know it's too late to block the commit locally. Maybe it does know that inof at the point one of those "commit time" hooks runs. I haven't tried it.

v/r, neal

Jeff King· Mar 21, 2012, 21:22 UTC · re: Neal Kreitzinger · lore

Re: Has anyone written a hook to block fast forward merges to a branch?

On Wed, Mar 21, 2012 at 04:14:21PM -0500, Neal Kreitzinger wrote:
Show 11 quoted lines
> >When I wrote my response, I assumed the intent was that people
> >_should_ be using --no-ff, but that he wanted to enforce it via hook
> >during a push to a central repository.
> >
> Would your script only work at "push time", or would it also work at
> "commit time"?  I'm not sure at what point he thinks he wants to
> enforce it.  I don't think git yet knows the info your script is
> asking for at "commit time" (ie, pre-commit hook and its brethren),
> and by the time it does know it's too late to block the commit
> locally.  Maybe it does know that inof at the point one of those
> "commit time" hooks runs.  I haven't tried it.

Completely untested, but you could probably do something like this in a pre-commit hook:

  if test "`git symbolic-ref HEAD 2>/dev/null`" = "refs/heads/master" &&
     ! test -e "`git rev-parse --git-dir`/MERGE_HEAD"
  then
    echo >&2 "Don't commit right on master. Go make a topic branch."
    exit 1
  fi

but I'm not sure if that would catch fast-forward merges or not. You might also have to do an ancestry check against MERGE_HEAD.

Whether such a thing would be useful depends on your workflow and exactly what you're trying to achieve. Usually people try to enforce policy like this on pushes, just because that is an easy central point (e.g., somebody who forgets to use topic branches or --no-ff may also forget to set up their hooks; but if you have a central publishing point, it's easy to set up the policies there just once).

-Peff

← back to recent threads