hi Yves and others,
I see two problems with pd-gui's handling of tcl stuff sent by the pd layer.
- The 20k limit can be hit very easily with array redraws, or
with toxy -- the more so, the bigger is the effect of glueing several commands, and sending them at once.
- The handler at the pd-gui side, pd_readsocket(), is not
reentrant -- some commands (like tk_getOpenFile) enter event loop, which might invoke the handler reentrantly, thus overwriting any additional commands that are glued to the end of the blocking command. However, I guess, making pd_readsocket() reentrant would have to involve some sort of stamping sys_gui data (with serial numbers?).
Basically, the weakness is of the pd-gui, not of the threading patch (which sometimes makes that weakness more evident, perhaps).
Krzysztof
Yves Degoyon wrote: ...
it's simply a question of communication between pd and pd-gui which happens either if it's threaded or not.