[STUMP] Interesting Bug...

[Brit Butler]

Stump-devs,

I'm a recent stump convert and very much enjoying it. I've stumbled upon a reproduceable bug but it's kind of weird. Bear with me...

I have a Lenovo X200 laptop with a dock, the dock is connected to an external flatscreen via VGA. With Xmonad I would get home, drop on the dock, use lxrandr to enable the second screen and be off to the races. This didn't work with stumpwm so I wrote the fix-layout defcommand in my .stumpwmrc which works handily. Connected to stump with slime, it's interesting to note that (current-screen) always returns the screen stumpwm first loaded with, even if I turn OFF the VGA monitor in lxrandr and turn on the laptop screen (or vice versa). (current-head) always shows the accurate resolution of the active screen. It's also interesting to note that *screen-list* never has more than one entry, including with both screens on in lxrandr.

So, the weird bug:

When I "flip" monitors in lxrandr and hit OK, stumpwm hangs. The window I'm in receives all keyboard input and, generally, it's like stumpwm isn't there at all. If I switch to a tty, open emacs, slime-connect to stump and try to run (loadrc) or most anything in the :stumpwm package, I get a backtrace like that seen in http://redlinernotes.com/docs/stumpwm_backtrace.txt. The brief absence of the former (current-head) causes something not to type-check as head and even though a new head takes its place, stump hangs. Aborting to SLIME's toplevel in the debugger returns stump to it's normal state. I'm happy to try and produce fuller debugging information. 

I'm not sure what a sensible fix would be but I'm open to suggestions. This may be a weird use case. If this is too esoteric to fix, I won't object. I could always try to find where to put the appropriate restarts for the type-error myself, time allowing...

It would be nice in any case if anyone could tell me how to add a *head-changed-hook*, so that I can have fix-layout run automatically when I switch heads.

My .stumpwmrc is posted at http://redlinernotes.com/docs/.stumpwmrc for the curious as is a screenshot (188k, full res) of the stumpwm slime session with (current-head)/(current-screen) calls.

Cheers,

Brit Butler