charm/hook: moved from worker/uniter

charm/hook: moved from worker/uniter

After a discussion, we realized it makes sense to move the
hook module from worker/uniter into charm/. All relation
types and global hooks are defined there.
In addition, I added a ScanCharm() call, which will be
used to fix bug #864164 and can be useful to validate
a charm's hooks dir against the metadata defined relations.

https://code.launchpad.net/~dimitern/juju-core/reorganize-charm-hook-consts/+merge/148436

(do not edit description out of merge proposal)

[dimitern, 2013-02-14 12:28:06]

Please take a look.

[fwereade, 2013-02-14 14:17:03]

As discussed online:

https://codereview.appspot.com/7303091/diff/1/charm/hook/hook.go
File charm/hook/hook.go (right):

https://codereview.appspot.com/7303091/diff/1/charm/hook/hook.go#newcode38
charm/hook/hook.go:38: )
A couple of const slices of UnitHooks and RelationHooks -- that can be used by
the putative Meta method such that we can add hooks to this file and have them
handled correctly -- would be nice.

https://codereview.appspot.com/7303091/diff/1/charm/hook/hook.go#newcode60
charm/hook/hook.go:60: type Info struct {
This should stay in worker/uniter/hook; it's only relevant to the uniter.

https://codereview.appspot.com/7303091/diff/1/charm/hook/hook.go#newcode95
charm/hook/hook.go:95: 
As discussed online, we don't (at the moment...) want anything from here down;
we just want a method on Meta to tell us what hooks might be valid, and check
files against its result in BundleTo/ExpandTo to know whether we should pay
attention to the executable bits.

https://codereview.appspot.com/7303091/diff/1/worker/uniter/relation/hookqueu...
File worker/uniter/relation/hookqueue.go (left):

https://codereview.appspot.com/7303091/diff/1/worker/uniter/relation/hookqueu...
worker/uniter/relation/hookqueue.go:6:
"launchpad.net/juju-core/worker/uniter/hook"
Having the constants in one and Info in the other will ugly up the code here and
elsewhere in the uniter -- maybe import them as hook.Install vs uhook.Info, or
something? I don't think this infelicity overrides the value of keeping them
separate.

[dimitern, 2013-02-14 16:34:07]

I'm closing this because of an lbox issue and will repropose it.

https://codereview.appspot.com/7303091/diff/1/charm/hook/hook.go
File charm/hook/hook.go (right):

https://codereview.appspot.com/7303091/diff/1/charm/hook/hook.go#newcode38
charm/hook/hook.go:38: )
On 2013/02/14 14:17:04, fwereade wrote:
> A couple of const slices of UnitHooks and RelationHooks -- that can be used by
> the putative Meta method such that we can add hooks to this file and have them
> handled correctly -- would be nice.

Done.

https://codereview.appspot.com/7303091/diff/1/charm/hook/hook.go#newcode60
charm/hook/hook.go:60: type Info struct {
On 2013/02/14 14:17:04, fwereade wrote:
> This should stay in worker/uniter/hook; it's only relevant to the uniter.

Done.

https://codereview.appspot.com/7303091/diff/1/charm/hook/hook.go#newcode95
charm/hook/hook.go:95: 
On 2013/02/14 14:17:04, fwereade wrote:
> As discussed online, we don't (at the moment...) want anything from here down;
> we just want a method on Meta to tell us what hooks might be valid, and check
> files against its result in BundleTo/ExpandTo to know whether we should pay
> attention to the executable bits.

Done.

https://codereview.appspot.com/7303091/diff/1/worker/uniter/relation/hookqueu...
File worker/uniter/relation/hookqueue.go (left):

https://codereview.appspot.com/7303091/diff/1/worker/uniter/relation/hookqueu...
worker/uniter/relation/hookqueue.go:6:
"launchpad.net/juju-core/worker/uniter/hook"
On 2013/02/14 14:17:04, fwereade wrote:
> Having the constants in one and Info in the other will ugly up the code here
and
> elsewhere in the uniter -- maybe import them as hook.Install vs uhook.Info, or
> something? I don't think this infelicity overrides the value of keeping them
> separate.

Done.