Bazaar dirstate locking (was: lock_* API change)

[Eli Zaretskii]

> From: "Stephen J. Turnbull" <stephen at xemacs.org>
> Cc: robertc at robertcollins.net,
>     bazaar at lists.canonical.com,
>     monnier at iro.umontreal.ca
> Date: Mon, 17 May 2010 22:52:15 +0900
> 
> Who's talking about VC?  I'm talking about Just Plain Scripting:
> 
> (defun bzr-commit-in-background ()
>   (interactive)
>   (find-file ".msg")
>   ;; Insert output in current buffer.
>   (shell-command "bzr ls -V -R --unknown" t)
>   (recursive-edit)
>   ;; If Emacs' `shell-command' doesn't support the trailing `&' itself
>   ;; and your shell doesn't either, use `start-process' or similar.
>   (shell-command "/usr/bin/bzr commit -F .msg &"))

And this is different from what I do now (run bzr outside Emacs) in
what ways?

VC is different, because it doesn't just run "commit", but does a lot
of other useful things as part of that.

> As for the other claims you make, I'm not an expert on bzr internals,
> I just know enough about databases that it seemed plausible to me that
> a lock might be needed in the current design.  The expert is Robert
> Collins who says he's pretty sure he has the dirstate locking breakage
> fixed, it was an unfortunate design mistake and not terribly invasive
> to fix.  Even though (until Emacs adopted Bazaar) users preferred to
> switch their workflows rather than wait or even test a fix, a fix is
> already in progress.  Seems to me like things are working here.

Sounds like we are in agreement here: this should be fixed.  A few
mails back, it sounded like you are blaming me (or Emacs) for using a
wrong workflow.