nspluginviewer priority on newer linux kernel

[James Richard Tyrer]

Thiago Macieira wrote:

> The problem might exist (I haven't tried 2.6.23), but the point is that 
> threads would not solve the problem. The plugins are already run in a 
> separate thread (actually, a whole separate process, but that's just one 
> specific kind of thread).
> 
> Your suggestion of solution is the actual implementation right now.
> 
I don't think so.  I notice that when Konqueror refuses to respond or is 
so sluggish that it is unusable that other things still work OK.  I am 
not a Kernel expert (I did read the MINIX book some years ago) but my 
best guess here is that Konqueror doesn't have a process waiting for 
input and the other stuff that responds is waiting for input (listening 
for it at a socket IIUC).

When Konqueror won't respond, I can go to KSysGuard and kill off some 
nsplugins and the problem is often fixed.  Next time, I will try niceing 
them to 19 and see if that cures the problem.

So, my idea is that if we could split off a small thread that listened 
for input, did the GUI, and told the large thread (or threads) what to 
do that this refusal to accept input could be fixed.

 From a fusibility perspective, the refusal to accept input is larger 
problem that it being slow in accomplishing the task requested.

-- 
JRT