comments on release candidate

[John A Meinel]

Erik Bågfors wrote:
> Congrats for getting out a release candidate. I'd love to see a 0.8
> release soon.
> Here are some random comments on the RC...
> 
> 1) metadir format not default?
> This should really change since it's needed for shared, repos, bound
> branches,checkouts, etc.  Most of us use them, and I think they have
> been tested really good.
> 

I would agree with that metadir should be the default. But I think there
is some concern about upgrade churn. Where because 1 person upgraded,
everyone else has to upgrade to keep working with the same person.

I think perhaps the best thing would be to have a configure entry that
would let you set the default format you wanted to use. Which lets it
become a policy decision, rather than mandated by the bzr people. We
could still set the default to metadir, but it would let people change
the default if they wanted to be more conservative. Though usually it is
best to use conservative defaults, because novices won't know what is
going on, and advanced users can figure it out.


> 2) knits still experimental.
> Are they not planned for 0.8? 0.8 was going to be default for at least
> a year I thought? What's keeping it back?

I would love to see them be enabled. But they really aren't ready yet. I
don't know if it is just waiting on a couple of merges, but I have a
branch that fails to convert.
And last time I tried it, 'bzr status' was slower with knits than with
weaves.

> 
> 3) bzr branch cannot branch TO a remote location
> This bothers me
> 
> See https://launchpad.net/products/bzr/+bug/35566

I think right now we might be able to support branching into a
repository without working trees. Just that the Repository/Branch code
uses Transport, but WorkingTree uses direct os.* or open() calls.

> 
> 4) Paths are always from root of tree,
> See https://launchpad.net/products/bzr/+bug/30159
> 

Having seen svn's only commit from this directory down in action, I
really prefer bzr's commit everything unless you supply '.'. It at least
gives you the option. And I think it is a safer default (you don't
forget to commit stuff because it wasn't in the current directory).


> 5) : [bagfors at zyrgelkwyt]$ ; bzr init
> bzr: ERROR: Already a branch: .. Use `bzr checkout` to build a working tree.
> 
> But I already have a working tree
> 

Yeah, it should realize whether or not there is a working tree.

> 6) bzr help init-repo
> example:
>     bzr init-repo repo
>     bzr init --format=metadir repo/trunk
>     cd repo/trunk
>     (add files here)
> 
> Bad example (see earlier mail)
> 
> 7) : [bagfors at zyrgelkwyt]$ ; bzr help branch
> usage: bzr branch FROM_LOCATION [TO_LOCATION]
> 
> Not clear that TO_LOCATION has to be local. Some information on how to
> create a remote branch would be good.

At present, 'bzr push' will create a remote branch (without a working
tree). I've never really agreed with that, but some feel 'bzr push' is a
'publish' command.

Anyway, doing this worked:
$ bzr init-repo ~/,xxx
$ bzr push sftp://localhost/~/,xxx/sl
Inventory ok.
       0 revision(s) pushed.

Note that there is still a bug where pushing/pulling/branching without a
working tree says that 0 revision(s) were pushed. (At least into a
repository)

> 
> 8) bzr help checkout
> Description of normal checkouts is under --lightweight, that's where I
> would expect to read about ligthweight checkouts, not normal..
> 
> 9) bzr help commit
> Talks about bound branches, but I can't find anything else about bound
> branches. Under "bzr help checkout", it's never talked about bound
> branches, just "normal checkouts". (which happens to be the same, but
> that's not said anywhere)

Well, I think we need to revisit our terminology. If we are now calling
a "bound branch" a 'checkout'. Then stuff like 'bzr bind' doesn't make
sense. Effectively 'bzr bind' transforms a standalone branch into a
checkout. (or a repository branch...)

So there is certainly something that needs to happen here.

> 
> 10) bzr init defaults to ., bzr init-repo does not default to .,
> MAYBE this is intentional and good? Just inconsistent I think

I think init-repo should default to '.'. I thought there was a patch to
support this.

I have frequently done 'mkdir foo; cd foo; bzr init-repo', just to be
told that I need to supply a path and 'bzr init-repo .' just says the
path already exists.

> 
> 11) bzr pull cannot create a new branch, bzr push can create a new
> branch. I expect them to behave the same.
> 
> See https://launchpad.net/products/bzr/+bug/35566

As mentioned above, *I* think 'bzr branch' should create new remote
branches, and 'bzr push' should not.

> 
> 12) bzr help
> the warning
> WARNING: This is an unstable development version.
>          Please keep backups.
> 
> Should be removed, since this is a RC, or do you want to keep that under 2.0?
> 
> Regards,
> Erik
> 
> 

John
=:->

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 249 bytes
Desc: OpenPGP digital signature
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20060416/6768c356/attachment.pgp