Rakudo: JVM build is busted

[AlexDaniel]

Error during the build: https://gist.github.com/AlexDaniel/a3ff8848f785bb1e8a02da21e93cdcd3

Regressed after this commit, but it probably just exposes a different issue (RT#126702?).

IRC discussion: https://irclog.perlgeek.de/perl6-dev/2018-04-22#i_16078190

[zoffixznet]

Reproed the crash: https://gist.github.com/zoffixznet/136495e58a008e8f5425d1292e6aefe2

[zoffixznet]

Tried to dig at it today.

Basically it fails here to typecheck a Hash against Associative and if you dump the debugName, you'll see that it's the newly added *% slurpy in protos that it's typechecking with.

I'm suspecting the cause is the same or similar to what causes R#1566 in that if something is not yet fully-composed, it fails typechecks. So I tried to re-compose that hash before doing the typecheck, but kept getting null pointer errors with my attempts.

The other thing I've tried is moving Associative role above traits in sources list (didn't help). Also reverted all the traits to use the old (|) protos (since with them we don't care much about keeping right count/arity) but that just made it blow up with prefix:<->.

The builds are too slow and now I'm also getting Unhandled exception: java.lang.OutOfMemoryError: Java heap space after adding a debug print even after bumping the limit to 400GB. So I'm giving up on this.

passes the torch to the next one in line

Good luck.

[zoffixznet]

pmurias++ fixed it in https://github.com/rakudo/rakudo/commit/1ae4d2d426

Confirming the fix fixes the issue on my box.