Path encoding problems not mentioned in docs

[James Westby]

When starting to version an existing project you will do:

  bzr init
  bzr add

if your encoding and the paths don't match such that you get a
UnicodeDecodeError similar when doing the second step there isn't
any helpful documentation for dealing with it.

A better error message would be great, but it can't be made to disappear
entirely, and so documentation on how to deal with it would be great.

This was observed while watching some new users try to start with bzr.

Thanks,

James

[Martin Pool (mbp)]

Thanks for the user-testing data.

I know sometimes docs are the best place for a stopgap for a bug, or
perhaps a better error message is. The error message, perhaps with a
url for more info, is more likely to actually be seen at the right
time.

What should we suggest they do though? Check their locale is set
properly, or maybe rename the files that don't match?

I'm not sure if a general message about unicode would be specific
enough to be useful but we could probably help for some specific
errors.

--
Martin <http://launchpad.net/~mbp/>

[James Westby (james-w)]

On Mon, 18 Jan 2010 06:57:55 -0000, Martin Pool <email address hidden> wrote:
> Thanks for the user-testing data.
>
> I know sometimes docs are the best place for a stopgap for a bug, or
> perhaps a better error message is. The error message, perhaps with a
> url for more info, is more likely to actually be seen at the right
> time.
>
> What should we suggest they do though? Check their locale is set
> properly, or maybe rename the files that don't match?
>
> I'm not sure if a general message about unicode would be specific
> enough to be useful but we could probably help for some specific
> errors.

I agree.

The current approach of raising an internal error isn't helpful though.
A more specific error would be great.

I think the best we can do with documentation is explain why the problem
usually occurs and give suggestions for how to fix it.

Thanks,

James

[Robert Collins (lifeless)]

> if your encoding and the paths don't match such that you get a
> UnicodeDecodeError similar when doing the second step there isn't
> any helpful documentation for dealing with it.
>
> A better error message would be great, but it can't be made to disappear
> entirely, and so documentation on how to deal with it would be great.

In principle we shouldn't care about the encoding accuracy of paths
above the tree root. Perhaps we should open a separate bug about that.

-Rob