[openbeosnetteam] Re: Eureka!!!

> David wrote:
> > The problem is that select never returns, so that's the problem.
> > Essentially select is still truly broken on R5 despite Philippe's
> > best efforts... The problem may be related to the accept stuff I'm
> > trying to debug though, so we may kill 2 birds with one stone.
>
> Still not honoring the timeout past?
> Grrr. Someone should try to add some printf() in the
> libnet/select.c:on_select_timeout(). I dunno why, but
> when empty it seem that it doesn't unblock/interrupt the kernel
> select() call. I'm curious to understand why!

It's not as simple as that :( Basically, we send the request, get data back
but never wakeup select. That's the issue I'm worried about and I think it's
the same one that stops accept working... Anyway, I've just got things
building as I wanted in userland so I can do more debugging now :)

>
> > Hmm, don't count on that being a reality :) KDL will be still be
> > visited!
>
> Absolutly.
> Don't we all like this friendly KDL, do we?
>
> On this very topic, kernel modules and drivers can add their own KDL
> commands: look at add_debugger_command() documentation.
> Maybe it would be a great idea to add one here and there, to be able
> to dump some internals stacks structs and variables after dropping in KDL?

Yep, maybe a really good idea! I was going to but then figured it was better
just to get on with fixing the other bugs :)

david

Other related posts: