Re: g_idle killer ...

[Maciej Stachowiak <mjs eazel com>]

Michael Meeks <michael@helixcode.com> writes:

> Hi there,
> 
> 	The attached patch adds a g_idle handler ( I feel this is
> preferable to a timeout ) that kills the process if it has failed to
> register the correct IID after a few idle loops. This can of course race
> and cause grief; hence there is a function to turn this off before it
> begins, however I feel it is a lot better than what was there.

Oops, I already replied to the previous version of this patch. Anyway,
I don't think using an idle handler like this is appropriate; number
of passes through the idle loop is a completely arbitrary and
unpredicatable condition. While a fixed timout is somewhat arbitrary
too, I think I would like it better. 

> 	Ok; it also adds a test case which is nice, and here is the
> crunch; run the test program and see you get the nice message and the
> correct exception back, _but_ then oafd goes and chews 99% of your CPU
> which melts to a pile of slag.
> 
> 	Quite why this should happen I havn't diagnosed; it is outside my
> immediate realm of expertiese. I imagine it is sitting polling file
> descriptors from applied guesstimation...
> 
> 	Regards,
> 
> 		Michael.
> 
> -- 
>  mmeeks@gnu.org  <><, Pseudo Engineer, itinerant idiot