[MERGE] Re: ProgressTask UserWarning branching stacked->unstacked

[Martin Pool]

2009/1/26 John Arbash Meinel <john at arbash-meinel.com>:
> This should be happening any time you push using the generic fetch code.
>
> Martin added some code, such that if you create a nested progress bar,
> and then call ".update()" on the parent, it issues the warning you saw.
>
> And it turns out that the generic fetching code was creating a nested
> progress bar that it never used.
>
> The attached patch should prevent the warning.

It's nice that the warnings are catching this kind of nit; it's much
better than having the whole program abort when it finds a small api
problem.  On the other hand if it's a longstanding problem, or if it
occurs as a knock-on effect of an exception, it looks bad to users.

I'm wondering if we should send warnings only to .bzr.log by default
in release candidates and final releases, unless -W is given.  (I seem
to recall discussing this before.)  I know there are cases where this
is not precisely what you want, eg in testing plugins, but it seems
like a fair tradeoff.

-- 
Martin <http://launchpad.net/~mbp/>