[PATCH 1 of 2 V2] util: add a way to issue deprecation warning without a UI object

[Yuya Nishihara]

On Mon, 10 Apr 2017 15:41:08 +0200, Pierre-Yves David wrote:
> On 04/09/2017 03:08 PM, Yuya Nishihara wrote:
> > On Sat, 8 Apr 2017 11:37:20 +0200, Pierre-Yves David wrote:
> >> On 04/08/2017 10:16 AM, Yuya Nishihara wrote:
> >>> On Fri, 7 Apr 2017 19:03:55 +0200, Pierre-Yves David wrote:
> >>>> On 04/06/2017 05:44 PM, Yuya Nishihara wrote:
> >>>>> On Thu, 6 Apr 2017 16:09:07 +0200, Pierre-Yves David wrote:
> >>>>>>> If dirty hack allowed, I would do something like the following:
> >>>>>>>
> >>>>>>>   # util.py
> >>>>>>>   def _deprecwarn(msg, version):
> >>>>>>>       pass
> >>>>>>>
> >>>>>>>   # somewhere ui is available, maybe in dispatch.py
> >>>>>>>   util._deprecwarn = ui.deprecwarn
> >>>>>>
> >>>>>> That is a diry hack. I would prefer we did not used it.
> >>>>>
> >>>>> Yeah, that is dirty and I don't like it. But I'm okay with it as long as
> >>>>> it is a temporary hack.
> >>>>
> >>>> If you think the dirty hack is worth the potential extra exposure, I'm
> >>>> fine with it.
> >>>>
> >>>> However, I'm confused about your usage of "temporary hack" here. Why are
> >>>> you using temporary?
> >>>
> >>> I suppose the hack will hopefully disappear with the vfs compat layer. I'm not
> >>> sure if a permanent one is necessary.
> >>
> >> I think this kind of needs will appears again in the future. So I would
> >> not flag the solution as temporary. I would have made the function local
> >> to the scmutil module.
> >>
> >> So I would rather have a long terms solution. What do you think?
> >
> > If we need less ad-hoc one, I would add a global flag to enable deprecwarn
> > and set it only by global configuration (i.e. ui, not lui nor repo.ui.)
> 
> This seems a bit hacky (I also have some concerns about possible of 
> deprecated function being created before the ui is. but I'm not sure 
> they are funded).

IMHO it's as bad as using environment variable. Both are a kind of globals.

> > And perhaps will remove ui.deprecwarn() in favor of util.deprecwarn().
> 
> ui.deprecwarn is much more configurable and manageable that the global
> one (based on Python deprec warning). I do not think we should drop
> `ui.deprecwarc` in favor of the other one.

It's more configurable, but less easy to use because ui is required.

Anyway, I don't think it's worth spending time for designing the deprecwarn
feature. I'm okay for any of the ideas in this thread.