Perl 6 - the future is here, just unevenly distributed

IRC log for #moarvm, 2016-09-08

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
01:48 ilbot3 joined #moarvm
01:48 Topic for #moarvm is now https://github.com/moarvm/moarvm | IRC logs at  http://irclog.perlgeek.de/moarvm/today
03:56 geekosaur joined #moarvm
03:59 geekosaur joined #moarvm
06:11 domidumont joined #moarvm
06:17 domidumont joined #moarvm
08:04 zakharyas joined #moarvm
17:02 domidumont joined #moarvm
18:09 pyrimidine joined #moarvm
18:20 dalek joined #moarvm
18:31 FROGGS joined #moarvm
18:36 domidumont joined #moarvm
18:52 TheLemonMan joined #moarvm
18:59 dalek Heuristic branch merge: pushed 20 commits to MoarVM/line_based_coverage_4 by timo
18:59 timotimo ^- just a rebase of the line_based_coverage_3 branch
18:59 timotimo not really merge-able as it interferes with inlining in some strange way :(
19:06 Ven_ joined #moarvm
20:07 lizmat joined #moarvm
20:12 timotimo Thread 3 bound to an attribute of an object allocated by thread 1
20:12 timotimo at <unknown>:1  (/home/timo/perl6/install/share/nqp/lib/Perl6/BOOTSTRAP.moarvm:add_to_cache)
20:12 timotimo i get a million cases of this when using a cross thread write log on a script that crashes with "zeroed target thread id in work pass"
20:14 zakharyas joined #moarvm
20:19 timotimo i can make this go away by removing an autothreading operation from Promise.Bool
20:20 timotimo i imagine we'll want to protect this cache with a lock perhaps?
20:27 timotimo the crash i have now is in find_best_dispatchee
20:27 lizmat joined #moarvm
20:27 timotimo well, not necessarily crash, but it shows up in the cross thread write log
20:30 jnthn Cross thread write log doesn't mean bug
20:30 jnthn Not automatically
20:30 timotimo OK
20:30 timotimo i'm not sure how to properly debug this program
20:31 timotimo i wonder if it gets better if i throw out the native calls entirely
20:31 timotimo cool, it still crashes. and very fast, too
20:33 timotimo interested in the script?
20:48 timotimo removing a sub that doesn't even get called makes it finish :D
20:48 timotimo that sub is allowed to be empty and have an empty signature
20:48 timotimo but it is also allowed to come near to the start rather than at the very end
20:49 TheLemonMan JIT bug?
20:49 timotimo /home/timo/perl6/install/lib/libmoar.so(MVM_args_proc_cleanup_for_cache+0x23)[0x7f6cabfda273]
20:49 timotimo ^- this appears in a backtrace of something stumbling over a corrupted heap
20:49 timotimo double free or corruption
20:50 timotimo nope, i can get it to crash even when the jit is off
20:50 timotimo and even when spesh is off completely
20:51 timotimo https://gist.github.com/timo/610b39f340db90a4f14eab97c11fa2c3
20:51 timotimo if you want to experiment
20:53 timotimo mandelbrot is the sub that seems to make things stable when it gets removed
21:06 TheLemonMan timotimo, jitbug :D
21:06 TheLemonMan it's jumping to 0x3BB0
21:09 timotimo how can it be a jitbug if it also explodes when the jit is turned off? :P
21:09 timotimo must be one script with multiple bugs, i suppose
21:18 timotimo keep me posted on what you find out, anyway
21:19 TheLemonMan I finally managed to make it crash with the Jit turned off
21:19 TheLemonMan MVM_SC_WB_OBJ: Assertion `!(obj->header.flags & MVM_CF_FORWARDER_VALID)'
21:19 TheLemonMan no interesting output from valgrind though
21:20 timotimo yeah, same on my end
21:20 timotimo maybe ASAN has something to say, or maybe hellgrind

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary