…ough in JDK.
ARM is now a supported platform (at least on Linux) so add it to the CPU_TYPE enum for FFI.
Simply need to check length before accessing elements.
…e raised *some* sort of Exception. (See http://bugs.jruby.org/6940.)
String#split fails after doing a regex on UTF-8 chars. The logic for splitting with a String mixed zero-based and begin- based offsets into ByteList's bytes, which caused it to fail when begin != 0. Modified the enc-aware character shift offset comparison to always be begin-aware.
returning a failure.
with a nil value, it should be ignored. (The nil key, on the other hand, raises a TypeError.) Also adds a spec.
create a tmp directory in a nonexistent directory.
Regression: ruby,java(thread),ruby exception path results in nilClass exception This is another case where removal of the $! setting we did before running exception-handling blocks appeared to be premature or incorrect. There appears to be some paths through exception handling -- perhaps only in the interpreter -- that fail to set $! properly beforehand. I have restored one of these sets, and the case given in the bug report now passes. We should get these cases rolled in; would appreciate if you can adapt them to our JI specs and send us a PR.
…ion. Fixes JRUBY-6926