Tim Rowledge <tim(a)rowledge.org> wrote:
> Has anyone ever done any testing to see just how many passes of compacting can
> be survived? If we checked the actual free space (perhaps the lowSpaceSignal
> state is better) and refused to give any to the fwdBlock table would we simply
> be thrashing a bit more, or doomed? Or should I just give up because nobody
> can be bothered to actually think about how to do things right for a change?
Oh don't stop now! If it does not go into the standard VM then let's
make a Does Things Right VM. There are plenty of people who would live
dangerously by using a safer dev environment. Err, you know what I
On this particular issue, I have long wondered why the low-space
threshold is still unusably low. Fancy tricks aside, it seems like a
no brainer to increase this to a few megabytes.
To look to the future a little, we probably should have a
second-level debugging approach which puts you in a new
morphic project. It is quite possible that the reason you ran
out of memory, is still going to be in effect if you
stay in the same project. Two errors in a row--of any
kind--should dump you into a new morphic project. It's not
as nice as debugging in the original project, but sometimes it is the
only way to survive. This is a general mechanism that applies to memory
problems as well as plenty of others. Consider, e.g., a morph with all
its variables set to nil.
While we're on the subject, I'd love to see someone take a shot at the
old idea of catching infinite recursions. If the stack gets deep, then
pop up a debugger and ask if everything is okay. As I recall, you had
some ideas for implementing this in a simple way but it never happened.
Recursion is one of the ways that we run out of space, but checking for
out-of-space is not the nicest way to handle it.
Overall, I strongly feel that Squeak should be a *safe* language. It's a great
property and it is in our roots. Right now, there are a lot of easy ways
to accidentally wedge your image.... and many of them have simple
solutions should anyone care to spearhead this.