[Vm-dev] Performance, Quality and Process [was Array new: SmallInteger maxVal]

[David T. Lewis]

On Wed, Oct 21, 2009 at 10:19:17PM +0300, Igor Stasenko wrote:
> 
> 2009/10/21 Eliot Miranda <eliot.miranda at gmail.com>:
> >
> > Hi All,
> > ???? ??I'm not happy with this fix and I'm not happy with the lack of process behind it. ??First there has been insufficient discussion of what the right behaviour is. ??Second, the fix David has written does lots of computation (shifts) to check a valid size request that could be pushed earlier at initialization time, which would allow e.g. a vmParameterAt:put: to modify the max allocation request size. ??Third, there is no review of fixes; we just put them out there.
> > I'm concerned about performance, code quality and a lack of process for agreeing fixes. ??But at the same time I don't want to institute a bureaucracy or slow down the pace of development. ??Do others share my concerns? ??What suggestions have you?
> >
> > One problem here is that Cog will introduce a huge raft of changes to the VM and to Slang, and so possibly the whole issue is moot. ??We'll face the issues as we try and integrate my Cog VM into the squeakvm trunk. ??But it might be worth thinking a little about the issues up front.
> > David, I know you're technical lead, and I'm not trying to depose or undermine you. ??But I do think we can benefit from discussion and review of major changes. ??Alas, my Cog work not being generaly available yet is going to cause problems down the line. ??I need to at least hurry up and get the StackInterpreter released.
> 
> Hi Eliot & Dave.
> 
> I think that Dave's fix is a quick way to close the security hole.
> I mean, it is good to have some critical issues closed quickly and
> deliver the 'hot' fix than having no fix at all.
> And surely, it should stay open for further discussion how to make it
> better/cleaner/faster/safer etc..
> 
> I hope nothing in this fix is unrevertable, which can't be changed in
> future versions.

Hi Eliot,

No worries, the changes are easily reverted and I definitely welcome
more review and better solutions.

There is a Mantis entry for this as well (http://bugs.squeak.org/view.php?id=7405),
currently in status "testing". The specific changes that I added on
SqueakSource are in a change set on the Mantis report.

Please don't think of me as technical lead; the "VM team leader" is
an editorial and facilitation role and I fully expect to defer to others
in technical matters.

Dave