Re: [Gnu-arch-users] Re: fixing and extending "selected commit"

[Robert Anderson]

On Sat, 2003-10-04 at 16:35, Miles Bader wrote:
> On Sat, Oct 04, 2003 at 10:16:03AM -0700, Robert Anderson wrote:
> > No, "you" is the user of tla.  Are you really missing this point
> > wholesale?  I find that hard to believe.
> 
> Perhaps you're doing a very bad job of making it...

I dunno, Tom got it immediately, and noted that it was "very well said."

> > In my opinion, there should be an invariant for commit/get pairs: that
> > any tree that is created with "get" must be a tree that existed when a
> > "commit" command was given.
> 
> OK, that's your opinion.  Personally, I think that's a reasonable way of
> operating generally, but it would be silly and pedantic to use tla to try to
> enforce the notion.

Of course.  That's exactly what I said.  Why you are trying to imply
that I said otherwise is a little baffling.

> > But what I am saying is that it is definitely a bad thing to have that
> > be the ONLY option.
> 
> What on earth does _that_ mean?

It means exactly what it sounds like.  That if there was one option for
committing partial trees, it should be one in which the tree states for
"get" and "commit" follow the invariant I suggested, as opposed to not
following it.  I'm not sure how I can explain that more plainly.

> You can always _not use_ restricted commits.

Oh, come on.  The issue here is: "what are the options for committing
only parts of changes to trees?"  At least, that's what I'm talking
about.  If you're talking about something else, spell it out.

> If there's a desire to add some automatic way of generating a coherent tree
> for a (theoretical) pre-commit hook to use for test,

Why are you so hung up on "pre-commit hooks"?  That is a separate issue,
although one that will follow simply if my criteria for partial commits
are followed.  I personally don't want or need "pre-commit hooks".

> > And that would suck.  If we're going to have one way to do it, it ought to
> > be the way that enables good habits, not the way that enables sloppy
> > habits.
> 
> So basically you're saying you wish to not have restricted commits at all,
> and force users who want the functionality to follow a more roundabout
> method to achieve them?

I said exactly what I said and nothing else.  I said if there were _one_
way to do "partial commits", in the sense that I have a tree, and
through some process only a portion of the changes are then included in
the next revision in the archive, it should be one which follows the
invariant principle I stated.

Of course, I don't think there should be one way.  I think there should
be multiple ways, and if one of them doesn't respect that invariant
that's fine.  I'm not asking (as you falsely implied) to use tla to
_enforce_ the invariant.  Only to make it _possible_ to follow it, if
such functionality exists.

> OK, here's my opinion:  No, that's silly.  You're perfectly free to use a
> pre-commit hook to enforce the notion for yourself of course (once they're
> implemented).

I'm really not understanding your "pre-commit hook" notion here. 
Perhaps you made this as a result of a misunderstanding which I've
cleared up.

Bob