Problems with "pull --update"

[Matt Mackall]

On Tue, 2010-01-05 at 13:50 -0500, Greg Ward wrote:
> I have discovered two problems with "pull --update": one quite
> trivial, one possibly deeper.
> 
> 1) I'm pretty sure the help is incorrect: it says
>       update to new tip if changesets were pulled
>     but shouldn't that be
>       update to new branch head if changesets were pulled
>    ?  Someone tell me I'm wrong, or I'll send a patch.  ;-)

Yep, that needs fixing.

> 2) If you're expecting the same behaviour as "hg pull && hg update",
> you're in for a surprise.
>     At least this is clearly documented, but it means that users
> coming from CVS or Subversion
>     cannot use "hg pull -u" to replace "cvs up" or "svn up".  In the
> general case, you have to
>     "hg pull ; hg update".  ;-(

It has always been thus.

> #2 only occurred to me because I am documenting our new workflow
> (we're moving from CVS), and to make things less painful, I was going
> to try to promote "hg pull -u" as the replacement for "cvs update": do
> it before you start working on a patch, and then again before you
> commit.  But it only works as long as you're already at a branch head.
>  If you happen to be at an older changeset and nothing gets pulled,
> then you start working at the wrong place.  ;-(  So it sounds like I
> have to document "hg pull && hg update" as the replacement for "cvs
> update".
> 
> If I had a time machine, I would just go back and change it so "hg
> pull --update" did the same as "pull && update", i.e. update
> regardless of whether anything was pulled.  But I guess that's against
> the compatibility rules.  Darn.  Anyone else annoyed by this?

Yes. Here's a crazy idea: add a --pull flag to update.

-- 
http://selenic.com : development and support for Mercurial and Linux