Re: Ebola! Impossible!

Andy> I have seen reports, and have now reproduced the scenario Andy> where the use of JDE results in errors concerning Andy> "old-assoc" looking at the code this can only occur if a Andy> bytecode file is ebolified, which seems impossible to me. Andy> Any clues as to what might be going on? I'm wondering if we Andy> still have bugs in excecute_optimized_program. I don't think so. My bet is memory corruption. I fixed one GC problem (the do_autoload patches) which was producing such errors. But recent GCCs have become very aggressive about optimization, and our code (especially --use-union-type=no) is vulnerable to that. It could also be bad code from the byte compiler. This hypothesis is supported by the fact that many of these reports go away if the reporter recompiles locally. Also, there was a lot of tweaking of the optimizer (especially by Martin) during 21.2, but I don't recall changes to the bytecode interpreter, which has been stable for a while. -- Institute of Policy and Planning Sciences http://turnbull.sk.tsukuba.ac.jp University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN My nostalgia for Icon makes me forget about any of the bad things. I don't have much nostalgia for Perl, so its faults I remember. Scott Gilbert c.l.py