Henrik,
A simple one would be to release the final Exceptions code as an update, and issue new VMs.
One reason why we're a bit hesitant to do this is that we are SOOOOOOO close to having a J3 for everyone to play with - as of today, j3 did run approx. 3922 bytecodes before locking up on MacOS, due to a disagreement between Ian and MacOS what the "right" stack frame is (of course, Ian is right and MacOS is wrong...)
Doing the exception modifications would require several changes in the j3 and right now we want to look at porting j3 to a number of key platforms. So I wouldn't bet on the exception modifications in the immediate future.
- Andreas
"Raab, Andreas" Andreas.Raab@disney.com is widely believed to have written:
Henrik,
A simple one would be to release the final Exceptions code as an update, and issue new VMs.
[snip]
Doing the exception modifications would require several changes in the j3 and right now we want to look at porting j3 to a number of key platforms.
Whilst it is true that the j3 code would need some changes (simple ones according to the discussions with Ian, but not essential to the immediate need to get the thing working) there is absolutely no need to have them done for j3 in order to let everyone else have the advantage of working unwinds/exceptions right now. The image changes are totally benign - you can run an updated image on an old vm and get exactly the same behaviour as before (ie wrong, but no crash etc) and you can run an old image on an updated vm (with the same results). Since the un-fixed j3 would count as an old vm for this purpose, I don't think there is any problem.
Indeed, if the intention of having the j3 as a plugin works as it should, an updated vm bytecode vm running an updated image should still be ok if the j3 plugin were present; it would simply revert to being 'broken'.
tim
squeak-dev@lists.squeakfoundation.org