@hamzaremmal Maybe you could review since you left a comment on the issue. Reading the doc for TryCatchPatterns i am still convinced that this is the expected behavior.

Hi @eejbyfeldt, the diff seems small but I want to double check things for a spec point of view before approving. I'll add it into my todos of next week.

I think this PR has the correct behaviour, given we consider the catch part as pattern matching in spec.

I'm wondering, can we generate byte code for simple catch like case _: E1 | E2 | E3 | ... directly as well, rather than generating another pattern match, as Java supports multi-catch?

I remember that try-catch doesn't support TypeTest but does support ClassTag. (I remember replying to someone's question on the forum.)

I have a local test about saferExceptions where a transparent inline method expands to try-catch but I didn't get it to work; my point is that try-catch isn't as simple as "if I can do it with a plain pattern match, it works in catch".

I have a local test about saferExceptions where a transparent inline method expands to try-catch but I didn't get it to work

I'm curious, can I take a look of the test?

@noti0na1 Sorry I have no idea why I wrote this. Probably just noodling.

import language.experimental.saferExceptions
import java.io.IOException
import scala.reflect.ClassTag

def io(): Unit throws IOException = throw IOException()

def ok =
  try io()
  catch case ioe: IOException => println("drat")

transparent inline def nok[E <: Exception : {CanThrow, ClassTag}](inline body: CanThrow[E] ?=> Unit) =
  try body
  catch case e: E => println("darn")

def test =
  nok[IOException]:
    io()