thread deadlock in BMenuField

[augiedoggie]

While attempting to diagnose a problem in the BeAE program, I came across some odd behavior within BMenuField. The problem only occurs, on average, around 1 out of every 5 tries. Basically you have a BMenuItem attached to a BMenuField, which when selected, destroys the BMenuField object. The BMenuField destructor then gets stuck in wait_for_thread() waiting for the mouse tracking thread to die. I hope I've made that clear :)

I'm attaching some sample code which might explain it better. I assume that this code and the stuff from BeAE worked on BeOS, but, I no longer have it installed to test.

[augiedoggie]

As a side note. I don't see the attached menu being deleted anywhere in the BMenuField source. Isn't it supposed to own the attached BMenu?

[augiedoggie]

Replying to augiedoggie:

As a side note. I don't see the attached menu being deleted anywhere in the BMenuField source. Isn't it supposed to own the attached BMenu?

Umm, yeah. Disregard that last comment. I wasn't thinking clearly :) I see the internal menubar and menuitem(s) are added as child views.

[jackburton]

Still reproduceable ?

[augiedoggie]

Replying to jackburton:

Still reproduceable ?

Yes, still locks up the application. Tested on x86_gcc2 - hrev48282

[pulkomandy]

I can't apparently reproduce the first problem (but then I'm not sure if the window is hung or not, as it's empty. It does close fine).

I uncommented the call to _CreateView in the attached file. In that situation, it seems that the menu from the newly created menufield opens under the window instead of over it (I can see it after closing the window).