Merge lp://staging/~mbp/bzr/557886-update-file into lp://staging/bzr

ps, i'm doing this only in 2.5 because firstly, it's something people can avoid by using the command carefully, and secondly it's not impossible there are shell scripts that do run against a subdirectory and cope with that behaviour, and I wouldn't want to break them in a stable series.

33 + If the tree's branch is bound to a master branch, bzr will also update
34
35 If the tree's branch is bound to a master branch, it will also update

Diif artefact or true duplicate ?

Overall, I'm pretty sure some people *really* want to be able to update their whole tree while being in a subdirectory.

May be you should add a check that dir != '.' for them ?

Voting needsnefixing for the typo, up to you for the suggestion about the additional check.

Ha crap, most important bit forgotten: Thanks for updating the doc and addressing a well known pain point.

On 28 September 2011 16:49, Vincent Ladeuil <email address hidden> wrote:
> Review: Needs Fixing
>
> 33      + If the tree's branch is bound to a master branch, bzr will also update
> 34
> 35      If the tree's branch is bound to a master branch, it will also update
>
> Diif artefact or true duplicate ?

A mistake, thanks for spotting it.

> Overall, I'm pretty sure some people *really* want to be able to update their whole tree while being in a subdirectory.

They can always say 'bzr upgrade' with no arguments.

> May be you should add a check that dir != '.' for them ?

well, it does handle '.' in the root of the tree, which turns into a
relpath of ''.

I don't think 'cd subdir; bzr update .' is good to support because
that might also be taken to mean upgrading just the subdir.

m

> > Overall, I'm pretty sure some people *really* want to be able to update
> their whole tree while being in a subdirectory.
>
> They can always say 'bzr upgrade' with no arguments.

Are you sure ?

>
> > May be you should add a check that dir != '.' for them ?
>
> well, it does handle '.' in the root of the tree, which turns into a
> relpath of ''.
>
> I don't think 'cd subdir; bzr update .' is good to support because
> that might also be taken to mean upgrading just the subdir.

I was talking about 'cd subdir ; bzr update' which used to work.

In other words, the 'dir' default value should be None so you can distinguish whether the user specified a 'dir' argument or not and apply your check only when the user did specify an argument and leave the previously working use case untouched.

If I read your patch correctly, people doing 'cd subdir ; bzr update' will get an error about trying to update part of the tree. I think it's a reasonable assumption for a user to expect 'bzr update' and 'bzr update .' to have different meanings and for 'bzr update' to indeed update the whole tree, 'bzr missing' and 'bzr pull' work in subdirs for example.

Hmm, 'bzr help update' while mentioning 'bzr update [DIR]' doesn't document what DIR means though...

> > > Overall, I'm pretty sure some people *really* want to be able to update
> > their whole tree while being in a subdirectory.
> >
> > They can always say 'bzr upgrade' with no arguments.
>
> Are you sure ?

Well, I added a test :-) and it turns out that you're correct, this does not work.

update' which used to work.
>
> In other words, the 'dir' default value should be None so you can distinguish
> whether the user specified a 'dir' argument or not and apply your check only
> when the user did specify an argument and leave the previously working use
> case untouched.

OK, that seems like a good way to fix it.

> Hmm, 'bzr help update' while mentioning 'bzr update [DIR]' doesn't document
> what DIR means though...

I'll address that too.

I've fixed the issues vila pointed out (thanks) and I think he verbally ok'd this.

sent to pqm by email

sent to pqm by email