I did some more test yesterday, and have noticed that when I plug the icd2, it doesn't get recognized ( and that was obvious ).
The worst thing is that when I read the content of the 4550 (with a programmer, not the icd itself ) to check if is still well programmed I see lot's of difference .
Basically it seems that when plugging the icd, something happens in the circuit that put the micro in programming mode and then something other change it's content.
This could happens if the pic goes in programming mode and there is some kind of noise on the RB used to programm it ( or the pins are left "floating").
Now I wonder why the pic goes in programming mode, and where there could be the leak that make it hot.
Any idea ?
Pigi
Added after 1 hours 26 minutes:
In effect there is no chance the 4550 goes in programming mode.
Anyway I have added a resistor to MCLR and reprogrammed again with the MCLRE enabled.
Now it works again, so I start to suspect some problem with my usb ports.
I'll keep you informed of any change.
For the moment thanks to all.
Pigi