Patch: jcf-write and last_bc

[Alexandre Petit-Bianco]

Per Bothner writes:

> Insert `nop' if `jsr' immediately follows `monitorenter'.
> 
> But there is nothing about 'jsr' in the patch itself.

The ChangeLog entry might be wrong :-(. The problem originally had to
do with emitting verifiable bytecode for empty synchronized blocks.

> The patch is for CASE CONVERT_EXPR etc.  What does that have to do
> with monitorentry+jsr?  At the very least we need a comment
> explaining what is going on.

There's one in the jcf-write.c hunk.

If there's a better way to detect this situation and emit what I
recall to be a necessary `nop', I'll in favor of getting rid of the
hackish last_bc trick I installed.

./A