Feature Request / Question

[Michael Diamond]

On Thu, Jan 6, 2011 at 6:08 PM, Greg Ward <greg-hg at gerg.ca> wrote:
>
>
> Although most people would argue that problem XYZ and problem 2 really
> should be separate changesets.  Just commit as you go.
>

I'd definitely agree with this sentiment.  If you find yourself wanting to
"commit" without committing, you probably ought to just commit ;)

I recently worked on a project with several Mercurial novices, and this was
the hardest thing for them to wrap their minds around, it really is ok to
make a lot of commits - and the more granular the commits, the easier time
you'll have looking back.  Similarly, they were often afraid to commit
because they weren't completely sure about what they'd done.  But really,
the only thing you need to be confident about is pushing, commits can be
bad, it's not the end of the world.  I screw this up occasionally, for
instance if I commit a bug fix too quickly, I'm tempted to rollback the
changeset and fix the fix.  Almost invariably, I discover I would have been
better off creating an additional commit instead of trying to alter history.
 I realize I actually could have used the rolled back revision, and I'm SOL.


Michael
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://selenic.com/pipermail/mercurial/attachments/20110106/8731b707/attachment.htm>