$ cat bug1.in
foreach value in { "Name": "Steve", "Location": "Finland" } {
printf("Value %s \n", value );
}
And:
$ cat bug2.in
if ( false ) {
printf("OK\n");
}
The issue here is obvious when running the code (note you must run the second with the optimizer disabled):
$ ./evalfilter run bug1.in
Value Finland
Value Steve
Failed to run script: instruction pointer is out of bounds
$ ./evalfilter run -no-optimizer bug2.in
Failed to run script: instruction pointer is out of bounds
The issue here is exactly the same as that resolved and described in :
We generate a JUMP instruction which jumps "after" the block. The problem is nothing comes next. In the past such a thing would have been masked by the (mandatory) return instruction.
Now we have to deal with it - we can generalize the previous solution, changing OpTrue + OpPop into OpPlaceholder which will be like OpNop, but not optimized away.
Two examples expose a bug:
And:
The issue here is obvious when running the code (note you must run the second with the optimizer disabled):
The issue here is exactly the same as that resolved and described in :
We generate a JUMP instruction which jumps "after" the block. The problem is nothing comes next. In the past such a thing would have been masked by the (mandatory) return instruction.
Now we have to deal with it - we can generalize the previous solution, changing
OpTrue+OpPopintoOpPlaceholderwhich will be likeOpNop, but not optimized away.