[RFC] Avoid format deprecation warnings on read

[John Arbash Meinel]

Aaron Bentley wrote:
> Hi all,
> 
> It's pretty annoying to be told to upgrade a branch when you don't have
> write access to that branch.
> 
> I'd like to propos that we only issue format deprecation warnings on write.
> 
> I would settle for not issuing format deprecation warnings on http,
> though of course, local files may also be readonly.
> 
> Aaron

lock_write() sounds about the right time to do it. It would fix my bug
where I'm using an old-format branch to block access to my repository
when I'm working in test directories. (I have a repo in ~/dev, and a old
format branch in ~/dev/tmp, so I can just cd ~/dev/tmp, and play around
with bzr behavior). But every time I do 'bzr init', it warns me that
"tmp" is an old format and should be upgraded.

However, this doesn't solve your 'bzr upgrade' issue. IIRC we still lock
the branch when we are about to upgrade it. So maybe something like what
we do with unsupported formats. Where we pass a special parameter to the
constructor, and then when lock_write comes around it won't try to warn.

But in the short term, probably lock_write() is better than the constructor.

John
=:->

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 254 bytes
Desc: OpenPGP digital signature
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20060911/f66d2ebc/attachment.pgp