[KDE/Mac] QStandardPaths on OSX and Windows (Re: An interesting read, perhaps...)

[René J.V. Bertin]

On Saturday April 23 2016 22:05:43 David Faure wrote:

> > IIRC it's not that the unittests don't run, it's that QSP on OS X doesn't have a special testing mode.
> 
> I don't see what makes you say that.

OK, so I should indeed have checked if memory served me well before sending that email. My bad (and frankly more preoccupied by those bad sectors on my Linux rig's HD :-/ )


On Sunday April 24 2016 08:26:27 Ben Cooksley wrote:

> > The best solution I can see until now is adding support (in QSP) for env vars ($QT_, not $XDG_)
> > and/or for defining paths in qt.conf (as the kde-windows guys started), both of which
> > would help OSX and Windows.
> 
> Last time I recall Thiago being absolutely adamant that the variable
> would have to be a XDG_ one not a QT_ one (despite the fact that XDG_
> vars are a fish out of water on anywhere but Linux/*BSD's).

I don't really agree with that. XDG may be a foreign protocol but that doesn't mean it cannot make sense in the context of cross-platform applications, just like DBus can be seen as a cross-platform desktop bus.

> Do you
> think we'll be able to get past him this time?

It is probably a good idea to present a QSP patch in the broadest possible context, so making as little reference to KDE and XDG as possible. One of the reasons our initial attempt to get a QSP patch in failed was that it was seen as too KDE-specific (and Thiago wasn't the most vehement opponent; I doubt he cares much about Mac or MSWin specific changes to QSP).

> Lack of time, and the machine's propensity to go to sleep on itself
> which we haven't been able to resolve has blocked any progress on CI

System Preferences/Energy Saver/Computer sleep -> Never ?
System Preferences/Energy Saver/Wake for network access -> tick


> In terms of Windows, if we restructure how we put together install
> prefixes to match how Qt looks for things we probably don't even need
> a qt.conf file there.

That should be valid everywhere, and joins my suggestion to initialise the various (ECM?) variables that dictate install locations via a utility like qtpaths .

Cheers,
René