finally block executed twice if exception throw in this block

Let me add a bit to its weirdness:

try {
	def x = 0
}
catch (Throwable t) {
}
finally { 
	println "in finally" 
	throw new RuntimeException() 
}

executes the finally block 3 times:

in finally
in finally
in finally
Caught: java.lang.RuntimeException
        at Try.run(Try.groovy:8)

I get the feeling this is related to let try-catch return a value.

So, taking the code example from my comment, if I a try block like "def x = 0" at the end of which instructions of the finally block get appended -> if the last statement of the finally does a "throw" instruction and after that we still do "return" instruction (say that is the mistake), do u mean that return instruction will still be reached after "throw"

I thought once it reaches "throw", it will switch execution path and not do the "return".

I have confirmed the issue has to do with how we generate the exceptions tables in bytecode. The include the inlined finall blocks, which they should not. But fixing this will require several changes in ACG, too much to do before 1.7.0 I think.

So I am scheduling this issue as a blocker for 1.7.1 and 1.8-beta-1

Is it really a blocker if it can be moved to a later release?

it is a blocker for that release menaing the release should not happen without that fix.

Shouldn't its priority be raised to blocker as mentioned in the comments - so that it doesn't get missed out before 1.7.1 / 1.8-beta-1?

Yup, feel free to up the priority to be sure we fix it for the next releases.

I am more interested in seeing this problem's solution actually. I want to see how current exception table entries are split up to exclude the exceptions thrown from the inlined finally block instruction ranges so that they don't get caught by the catch blocks.

Higher the priority, more the chances of me seeing the solution soon , but I will let Jochen raise the priority, if he wants.

should be fixed now... it was only a matter of excluding the catch for the finally

Roshan's example still runs finally twice for me

Since, in my example, the finally block is still getting executed twice, I am re-opening this issue.

Bump fix version due to reopen

add code tags

Just checking on the status of this. It seems to be fixed in trunk and 1_7_X (I assume merged) but not in 1_6_X?

I fixed it on trunk and 1.7.x. The files differ too much to port the fix to 1.6