Stéphane Ducasse wrote <stephane.ducasse@inria.fr>:
Something that we prepared since months one little step after the other is to be able to avoid polling events but to use a semaphore. This is not use in image but can be activated using
InputEventSensor installEventSensorFramework
Remember that VisualWorks did that in 1998 :).
Now we went over the 100 of Sensor direct references in the code and fix them. We should still analyze the ones that are left around 70. The idea is that sensor should not be invoked from Morphic since Morphic is handling VM event. So code that predates Morphic or that is outside morphic and that implement a loop can react strangely. So a lot of cleaning in perspective. Now on this road with fixed (not integrated yet) the raw VM event manipulation done at the border between Morphic and none Morphic world. We are getting there.
Now Janko I do not know. May be others will have ideas.
Try InputEventSensor installEventSensorFramework even if I do not see why this would change something.
I will sit with igor this afternoon to check the bug we found because some events were blocked on a waitQueue. The cool thing is that for this bug is that we can systematically crash our image.
Stef
Happy new year! Tnx for confirming, what i have been complaining last year! :-) Btw.: Do you consider remote events while bundling images, e.g. via s2s? Should JS Smalltalk like Amber being able to run Morphic one day, it could be very interesting to use Pharo via simple Browser even in multiuser mode. I think, this design approach would give Pharo a tremendous advantage over ALL other competitors... Have a nice time hacking! Guido Stepken