branch: Repository KnitPackRepository is not compatible with repository KnitRepository

[John Arbash Meinel]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Ben Finney wrote:
> John Arbash Meinel <john at arbash-meinel.com> writes:
> 
>> Talden wrote:
>> |>  Is it unreasonable of me to expect that *any* format supported in a
>> |>  Bazaar release (despite "no-one was supposed to use" it) should be
>> |>  upgrade supported in later versions?
>> |
>> | If a given format were labeled as experimental then yes I think the
>> | expectation of that support would be excessive.
>> |
>> | I think it's great that the opportunity to try new formats is
>> | available (for users planning a future migration and for the project
>> | needing more exposure for testing) but for those formats, the project
>> | shouldn't be burdened unconditionally with complete support.
> 
> "Here it is, but don't use it" is a rather dubious way of saying "not
> supported".
> 
>> | That lack of support should also, of course, not burden users by
>> | failing to communicate to them the experimental nature of a given
>> | option.
>> |
>> | Did the help for 'init-repo' clearly explain the experimental nature
>> | of the format in that version?
> 
> I don't know, I didn't use 'init-repo' to get that format. It was,
> IIRC, a 'bzr upgrade --format=FOO' that was recommended to me by
> (someone I took to be) a member of the Bazaar project.
> 
> Now, of course the caveats about executing commandlines handed out via
> IRC apply here. I would argue just as strongly that, if a format isn't
> intended to be used except for experimentation, it shouldn't be part
> of a general release.
> 
>> Actually, the format was not listed by 'bzr help init-repo' it was a
>> hidden format that the user would have to have other knowledge to
>> even know that it existed.
> 
> That "other knowledge" came in <URL:irc://chat.freenode.net/#bzr> in
> response to a request for help that is now long-forgotten.
> 
> I'm not looking to cast blame, but merely to point out the negative
> and, in my case, enduring consequences of having a "please don't press
> this button" function in the released product. Even if I have no
> desire to press that button, it can easily happen by accident.
> 

Well, my IRC log talks about it at the end of February:
http://bzr.arbash-meinel.com/irc_log/bzr/2008/02/bzr-2008-02-26.html
http://bzr.arbash-meinel.com/irc_log/bzr/2008/02/bzr-2008-02-27.html

At the end of 02-26 'bignose' is saying:
23:49:08	bignose	I'm getting an error <URL:http://pastebin.ca/919766>
trying to 'bzr upgrade'
23:49:32	bignose	the upgrade is prompted by a 'bzr commit' giving
"Repository KnitPackRepository('sftp://bzr@fs/%7E/.bzr/repository/') is
not compatible with repository
KnitRepository('file:///home/bignose/Projects/.bzr/repository/')"

The pastebin is gone, so I don't know what it was saying.

But it seems like someone else had used the flag, which caused you to
need to upgrade.

Unless it was even earlier. But "grep bignose.*upgrade" doesn't return
any other hits. Other than a question about not having a working tree
along with your repository.

I probably have some gaps in my IRC coverage, but I don't find anyone
recommending that you upgrade to --rich-root-pack or --dirstate-subtree
since the beginning of 2007.

John
=:->


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFIDUkBJdeBCYSNAAMRAvWsAKDULIqhl9I+RL/oMYXc1cirhh8/CwCfS7Fx
rki35ZJT+cp3cxkg5BoC55o=
=3rLH
-----END PGP SIGNATURE-----