Already running instances of rakudo crash during rakudo upgrades

[jkramer]

The Problem

When upgrading rakudo, rakudo processes started beforehand crash (and dump core) with a "bus error":

# time perl6 very-long-running-process.p6
zsh: bus error (core dumped)  perl6 very-long-running-process.p6
perl6 very-long-running-process.p6   16405.76s user 29.48s system 174% cpu 2:36:35.94 total

Expected Behavior

Presuming the script only did calculations in memory and didn't access and files/modules/... that could be modified/removed/overwritten during the upgrade process, the interpreter in memory should be independent from the binary etc. on disk so it shouldn't crash.

Actual Behavior

Crashes with a bus error (see above).

Steps to Reproduce

Have a script that runs for a long time (mine was doing lots of computations, although I'm not sure if this is necessary in order to reproduce the crash). While it's running, upgrade rakudo:

cd ~/rakudo && git checkout master && git pull && git checkout $(git describe --abbrev=0 --tags) &&   perl Configure.pl --gen-moar --gen-nqp --backends=moar &&   make && make install

Environment

• Operating system: Linux 4.11.2-1-ARCH
• Compiler version (perl6 -v): This is Rakudo version 2018.02 built on MoarVM version 2018.02 (after the upgrade)

[jkramer]

I did some more testing to reproduce the problem. First, a perl6 -esleep or similar doesn't suffice, apparently it only crashes when it's actually doing something. I used this line for testing: perl6 -e'my $x; loop { $x++ }'

The command (from make install) that causes the crash seems to be this one:

cp -- dynext/libperl6_ops_moar.so $HOME/rakudo/install/share/perl6/runtime/dynext

Executing this command manually in the rakudo checkout directory (after a successful make) reproduced the crash for me every time. Hope this helps.