On 08/22/2017 04:53 PM, Claes Redestad wrote: > Extending the existing StringConcatFactoryInvariants test we can easily get > most basic > combinations covered, along with some basic verification. > > This uncovered what appears to be a similar issue in the BC_SB SCF strategy, > which can be > dealt with by coercing the possibly non-String constant to String at > generation time:
(Sigh) The trouble is LDC, right? 986 Object cnst = el.getValue(); 987 mv.visitLdcInsn(cnst); 988 desc = getSBAppendDesc(cnst.getClass()); Coercing to String is fine here. > http://cr.openjdk.java.net/~redestad/8186500/jdk.01/ Still think testing for {null, Class, MethodHandle, MethodType} would cover more interesting corner cases. Thanks, -Aleksey
