'parrot -O2 oofib.pir' errors out, when -O1 succeeds

[coke]

Originally reported as  http://rt.perl.org/rt3/Ticket/Display.html?id=49968.

$ ./parrot examples/benchmarks/oofib.pir
fib(24) = 46368 0.298038005828857s
$ ./parrot -O1 examples/benchmarks/oofib.pir
fib(24) = 46368 0.298038005828857s
$ ./parrot -O2 examples/benchmarks/oofib.pir
Class 'B' not found

Still occurring in

[coke]

Still occuring in r47566

[cotto]

parrot's -O2 is known to be pretty broken. Perhaps we should deprecate it, make sure that it spits out some noisy warnings or otherwise make sure it doesn't trip people who expect it to work up.

[coke]

Then lets make O2 do the same as O1 for now (in case anyone is using it). we can leave the option in for now and then it later when we have working code.

[jkeenan]

Replying to coke:

Then lets make O2 do the same as O1 for now (in case anyone is using it). we can leave the option in for now and then it later when we have working code.

coke, cotto:

Do you want to take any particular action with respect to this ticket?

Thank you very much.

kid51

[bacek]

It's imcc optimizer actually.

[cotto]

I have a dim view of most of imcc's optimizations. My recommendation would be to either rip out any broken optimizations or just to tell people that non-default optimizations are known to be incorrect.

[jkeenan]

Replying to cotto:

I have a dim view of most of imcc's optimizations. My recommendation would be to either rip out any broken optimizations or just to tell people that non-default optimizations are known to be incorrect.

Can we get an update on the status of this ticket? Would the problem described have been addressed by the recent work on IMCC isolation?

Thank you very much.

kid51

[bacek]

-O2 optimizations if IMCC were removed long time ago. Closing ticket.