← Back to team overview

dolfin team mailing list archive

Re: Introduce a new dependency (QT)?

 

Sounds good.

--
Anders


On Wed, Aug 29, 2012 at 09:53:11PM +0200, Joachim Berdal Haga wrote:
> Right! I'll make it optional, probably with some functionality missing
> without QT.
>
> -j.
>
> On 29 August 2012 17:48, Anders Logg <logg@xxxxxxxxx> wrote:
> > On Wed, Aug 29, 2012 at 09:52:51AM -0500, Andy Ray Terrel wrote:
> >> On Wed, Aug 29, 2012 at 4:55 AM, Garth N. Wells <gnw20@xxxxxxxxx> wrote:
> >> > On 29 August 2012 10:36, Anders Logg <logg@xxxxxxxxx> wrote:
> >> >> On Wed, Aug 29, 2012 at 11:20:19AM +0200, Joachim Berdal Haga wrote:
> >> >>> On 29 August 2012 11:11, Garth N. Wells <gnw20@xxxxxxxxx> wrote:
> >> >>> > On 29 August 2012 09:42, Joachim Berdal Haga <jobh@xxxxxxxxx> wrote:
> >> >>> >> I would like to switch the window handling and event loop to QT,
> >> >>> >> because it's much more flexible and mature (stable) than VTK's. An
> >> >>> >> example of things that are hard to get working right with the VTK
> >> >>> >> window handling is to close a single plotting window.
> >> >>> >>
> >> >>> >> This will introduce a new dependency for plotting (in addition to
> >> >>> >> VTK). It will be optional, and if it's not configured then it's only
> >> >>> >> plotting that is disabled.
> >> >>> >>
> >> >>> >> Any protests?
> >> >>> >>
> >> >>> >
> >> >>> > I'd rather not have it as a dependency. I don't really want a major
> >> >>> > dependency for lightweight plotting. I think we should bear in mind
> >> >>> > that we have ParaView, MayaVi, etc for making 'real' plots, so the
> >> >>> > DOLFIN plotting should remain as simple as possible.
> >> >>>
> >> >>> I agree with keeping it simple. The reason I want to introduce it is
> >> >>> not to introduce anything complex, but to gain more robust window
> >> >>> handling / event loop. However: Opposition noted -- would option 2
> >> >>> (basic support for VTK-only) be acceptable to you?
> >> >>
> >> >> I don't see why QT would be a problem. Isn't the VTK dependency just
> >> >> as heavy?
> >> >
> >> > Are you saying the 2 x 'heavy' is the same as 1 x 'heavy'?
> >> >
> >> > I've just checked, and QT is a 229MB tar ball!
> >> >
> >> >> Or are there systems where VTK is easily available but QT is
> >> >> not?
> >> >>
> >> >
> >> > There are lots of systems where neither is available. Needing two
> >> > makes the configuration and build ever more complicated. I know
> >> > first-hand that our config and build needs work on non-Ubuntu/Debian
> >> > systems, which I'd rather have sorted out before adding big
> >> > dependencies.
> >> >
> >> > Garth
> >>
> >> Piping in from a guy who builds too much code on supercomputers, I
> >> would suggest not making QT a hard dependency.  It is a pretty intense
> >> build and takes lots of space.  We only support it on a few of our
> >> machines but all our machines support VTK.
> >>
> >> Although Joachim is right about the event loop.  QT is the best gui
> >> interface out there, but perhaps it should be more of an optional
> >> dependency.
> >
> > ok. My assumption here was that plotting was mostly interesting on
> > desktop machines (not compute servers) and QT is easily available on
> > desktops.
> >
> > But perhaps a bigger problem is packaging for Mac and Windows which
> > would require bundling QT.
> >


References