Will Coleda wrote:
On Mon, Dec 15, 2008 at 11:05 AM, Kevin Tew <t...@tewk.com> wrote:
Will Coleda wrote:
3) What's the status of ticket "-r33351 causes tcl segfault" (TT#10)?
Is this still an issue? Is anybody actively working on it? Are we
realistically going to get this issue resolved before tomorrow
evening?
Unless someone has specifically addressed the issues listed in the
ticket (And then failed to update the ticket), this is still open. I
don't have remote shell access due to the recent power outages, but
can double check this evening.
Thanks to chromatic, we know what the problem is.
In short:
Subs need to use ATTRs
The fix is pretty invasive, its NOT going to get fixed pre-release.
Can we back out the change that introduced the problem instead?
The Sub PMC implementation is inherently broken.
I think the better way to think about my subid changes isn't that they
introduced the problem, but that they just exposed a pre-existing problem.
I don't have a full understanding yet, but essentially parrot objects
and Sub PMC's don't work together correctly under inheritance.
The Sub PMCs don't use ATTRs so you can't create a Parrot Class that
inherits from them safely.
The fact that it has worked up till now has to do with by chance memory
layout and allocation patterns that have hidden the inherent problem.
It has been pure luck that its worked this long.
Any change in memory layout or allocation patterns of Sub PMCs has the
possibility of exposing/re-exposing the inherent bug.
So reverting the subid changes isn't a fix.
I didn't have the time this weekend to look at the problem.
I'll try again tonight, time permitting.
Kevin
I've looked at the problem, but don't have a patch yet.
Kevin
