Re: [RTC/PATCH] Add 'update-branch' hook

[Felipe Contreras <felipe.contreras@gmail.com>]

Junio C Hamano wrote:
> Felipe Contreras <felipe.contreras@gmail.com> writes:
> 
> >> >> I have a branch which should always be recompiled on update;
> >> >> post-update-branch would be a good place for that.
> >> >
> >> > And why would pre-update-branch not serve that purpose?
> >> 
> >> Because the code that needs to be compiled is not yet in the workspace
> >
> > And it won't be in 'post-update-branch' either.
> >
> >  % git checkout master
> >  % git branch feature-a stable
> >  <- update-branch hook will be called here
> >
> > The hook will get 'feature-a' as the first argument, but the code in the
> > workspace would correspond to 'master'; the checked out branch (pre or post).
> 
> The whole point of a pre- hook is to run _before_ the externally
> observable state changes due to the operation.
> 
> If Stephen has a separate build-tree that fetches from the branch
> every time the tip of the branch changes in this repository to
> produce build artifacts for the branch to be shared in his network,
> perhaps via NFS or something.  "git fetch" that will be run from
> that build-tree repository will *not* see the tip of the branch, and
> running such a hook will not be possible from a pre-update-branch
> hook.
> 
> We can certainly argue that such a hook could instead push to the
> build-tree repository using the commit object name,

Exactly, it could do that.

> but I tend to think such an argument is merely sidestepping the real issue.

So you grant that there is no reason anybody can think of why we would ever
want a post-update-branch?

> Some hooks do want to observe the state _after_ the operation [*1*], while
> some hooks can do without seeing exactly the state after the operation.

Yes, and when the operation is updating a branch, nobody can think of why we
would want the former.

> So while I am generally not very supportive towards post-anything
> hook, I would reject a claim that says "pre-anything can be used
> without inventing post-anything---do the same thing and allow the
> operation and you are done".  That is not simply true.

Let's make a bet, we go for 'pre-update-branch' and five years from now, if
there's no 'post-update-branch', you will publicly accept thta I was right.

Deal?

-- 
Felipe Contreras