[Haskell-cafe] GHCi with wxhaskell-dev

[Jean-Marie Gaillourdet]

Hi, 
On 10.01.2012, at 19:05, Dave Tapley wrote:

> I'm working on a dev branch of wxHaskell [1] and one of the new
> features is that you can use wxHaskell in GHCi (again), but I seem to
> have broken something in the process:
> Firstly: wxHaskell works in GHCi insomuch as:
> You can load the hello world sample [2] in GHCi, do "> main", see the
> window spawn, close it and be returned to the GHCi prompt.
> 
> What I'm interested in is the case where you call wxHaskell's start
> function [3] from GHCi.
> The result is becoming stuck with no way back to the GHCi prompt
> without resorting to sending kill signals.
> 
> I made a screen cast to illustrate my findings [4]:
> 1. Once you've done "> start $ return ()" you can press ^C all you
> want, there's no way back to the GHCi prompt.
> 2. If you send three (less has no effect) kills to the GHCi process it
> will exit, but it will exit GHCi as well (instead of putting you back
> at the GHCi prompt).
> 3. If you do "> start $ return ()" but *do not* try and ^C, you can
> send a single kill to GHCi and be returned to the GHCi prompt.
> 
> My questions:
> 1. Why doesn't GHCi respond to the ^C?
> 2a. Why does entering ^C from within GHCi then prevent you from
> getting back to the GHCi prompt?
> 2b. Why doesn't any other input (i.e. entering anything but ^C) cause
> this behaviour?
> 3. If you have sent ^C from GHCi, why do you then have to kill three
> times to get out of GHCi?
> 
> Feel free to pull the repo [1] and try it out, you'll need the latest
> dev release of wxWidgets [5] installed.
> 

I didn't try to reproduce your problem, but are you aware of the  -fno-ghci-sandbox flag? E.g. it helps with GLUT based programs on OS X. 

Cheers,
  Jean