Re: [Gnu-arch-users] [BUG] apply-delta target directory

[David Allouche]

On Thu, May 27, 2004 at 01:35:40AM -0400, Aaron Bentley wrote:
> David Allouche wrote:
> >On Tue, May 25, 2004 at 08:52:32AM -0400, Aaron Bentley wrote:
> >>David Allouche wrote:
> >>
> >>>"tla apply-delta REV-A REV-B" applies the delta changeset to the
> >>>current directory, not the tree-root.
> >>
> >>Yeah, I thought that was odd behaviour too.  I wasn't sure if there was 
> >>some reasonable reason for behaving this way, though of course,
> >>scripts like aba change-version can avoid this by manually
> >>specifying the tree root.
> >
> >
> >You can often work around bugs. It's better to fix them.
> 
> If they're bugs.

In my understanding, that is a bug.


> >>It comes down to the two-faced nature of delta.  If you run delta in a 
> >>non-Arch tree, it will compare to the current directory.  If you run 
> >>delta in an arch tree, it will compare to the tree-root.
> 
> >But in any case, using non-Arch trees with delta does not make sense in
> >Arch as I understand it. Please educate me if I'm wrong.
> 
> Actually, if you do "tla delta $dir1 $dir2 $dir3" where $dir1 and $dir2 
> are not project trees, it's pretty much the same result as "tla 
> changeset $dir1 $dir2 $dir3"

Where $orig and $mod are directories, the only reasonable behaviour for
"delta $orig $mod $dest" is to behave as "changeset $orig $mod $dest".

Let me rephrase: "in any case using non-Arch trees for computing
changesets does not make sense in Arch as I understand it".

> >>Given that behaviour, I guess it's reasonable to expect apply-delta to 
> >>behave the same way.  If people really want to apply the changeset to 
> >>the wrong directory, they can use delta and apply-changeset.
> >
> >
> >I cannot understand what you mean. You seem to say: "given <broken>,
> >it's reasonable to expect apply-delta to <be-broken>. If people want to
> ><break> then can <not-use-apply-delta>" which is contradictory.
> 
> Since delta uses the specified-directory-tree-root, not the specified 
> directory, it's reasonable to expect apply-delta to use the 
> specified-directory-tree-root, not the specified directory.  If people 
> want to use the specified directory, they can use delta to create the 
> changeset and apply-changeset to apply it to the specified directory.

I do not understand why you need such a reasoning, but at least we agree
on the conclusion.

> >In any case, I'd expect the changeset-applying function to shout and cry
> >if the user is trying to apply a changeset involving patchlogs to a
> >directory which is not a tree-root.
> 
> It's just doing apply-changeset.  It's not particularly smart.  It 
> doesn't know anything about patchlogs and what they might signify.

I think "apply-changeset" fits the description of "the
changeset-applying function" pretty well.

For architectural reasons, I can understand why the internal function
might want to know nothing about patchlogs, but still I do not
understand why the apply-changeset CLI (or any other incarnation of
changeset-applying functions) should silently allow such behaviour.

-- 
                                                            -- ddaa