CI Segfault when return type of KafkaSinkEncoder implementation is a union

[jtfmumm]

If we use the following code in celsius-kafka:

primitive FahrenheitEncoder
  fun apply(f: F32, wb: Writer): (Array[ByteSeq] val, (Array[ByteSeq] val | None)) =>
    wb.f32_be(f)
    (wb.done(), None)

then though it compiles on OSX and Dipin's Linux VM, it causes a segfault on Travis.

This signature does not trigger this:

primitive FahrenheitEncoder
  fun apply(f: F32, wb: Writer): (Array[ByteSeq] val, None) =>
    wb.f32_be(f)
    (wb.done(), None)

It may be that bringing this change over to Sendence compiler will address this:
ponylang/ponyc#1615

[SeanTAllen]

@jtfmumm did you mean #1615? that seems not like the right one.

[jtfmumm]

@SeanTAllen Dipin was dealing with this issue earlier and thought it might be related. @dipinhora

[dipinhora]

@SeanTAllen I thought #1615 might be the right one because what I encountered while developing the KafkaClient is that the AST sometimes causes either a segfault or some other not correct/useful error but which goes away once the AST is changed without necessarily changing the logic implemented. At the time, I didn't investigate because I didn't want to distract myself and it hasn't occurred recently for me so I can't try out #1615 to confirm if it fixes it or not.

I suggested #1615 to @jtfmumm because I assumed this issue in Travis was similar to what I was encountering but it is possible that the assumption is incorrect.

[SeanTAllen]

The first code listed isn't legally valid. The 2nd is but I am still getting a segfault.

[SeanTAllen]

#1190 which has ponylang/ponyc#1615 in it ran fine for CI. Closing this.

[jtfmumm]

There was a typo in the first code listed. Fixed and reopening

[SeanTAllen]

I thought I had a minimal example for this, but I don't.

[dipinhora]

@jtfmumm Why was this reopened? I thought the segfault was because of the bug fixed by ponylang/ponyc#1615.

[jtfmumm]

I thought you were saying this was fixed on mainline ponyc. Did I misunderstand?

…

On Fri, Aug 25, 2017 at 3:46 PM, Sean T Allen ***@***.***> wrote: I thought I had a minimal example for this, but I don't. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1178 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AEgtzE-MoZkxRHK4lf8m1pgGIxh078A9ks5sbtAwgaJpZM4PASOO> .

[SeanTAllen]

@dipinhora it was never tested for that change. i am testing it now.

@jtfmumm i thought i had a minimal example. i was tired and wrong.

[SeanTAllen]

This was fixed by #1615.