Fork me on GitHub
#cljdoc
<
2022-11-30
>
mpenet10:11:10

Is there a way to have the job run with java19 with --enable-preview? I am hitting an issue with a (somewhat bleeding edge thing) https://app.circleci.com/pipelines/github/cljdoc/builder/37933/workflows/d69459b7-cb6a-495f-b9ba-23d8a0867d38/jobs/54307

lread13:11:09

Thanks for asking @U050SC7SV. No, not yet. Production does API analysis currently only on jdk17. That's an interesting note on jdk19. I guess folks would typically want to use the --enable-preview java opt for jdk19?

mpenet13:11:35

likely yes, if you're using j19 it's probably to test the new features (like loom)

seancorfield17:11:07

At work, we've talked about adding --enable-preview to all our JDK invocations so that we can leverage preview fixtures in any JDK we use, so that might be worth considering regardless of the JDK used for running the analysis. At least you'd be future-proofed for any JDK upgrades that cljdoc gets 🙂

mpenet10:11:25

I looked at the docs and I don't think it's supported, but maybe I missed it

lread13:11:44

Nope, you didn't miss it, only jdk17 currently.

mpenet17:11:45

Yeah I suspected that :/

mpenet17:11:28

I missed that issue on gh, interesting discussion

seancorfield17:11:57

@UE21H2HHD What analysis would break switching from JDK17 to JDK19? (I hoped that link would give specifics -- or do I need to dig deeper into those linked issues in that comment?)

seancorfield17:11:55

Ah, the Thread/sleep issue? Yes, we ran into that updating to JDK19, but that was a while back and I'd forgotten.

lread17:11:11

Ya @U04V70XH6, that's the culprit I know about.

lread17:11:58

That issue is so old and long. Here's the https://github.com/cljdoc/cljdoc/issues/275#issuecomment-1326642637 on JDK19 Thread/sleep

lread20:11:22

Now that I have access to cljdoc's logs via Sentry (recently upgraded thanks to @martinklepsch and Sentry), this means I can easily see any unexpected exceptions being thrown when folks are using cljdoc. I'll start looking at these sometime soon. This should help to improve the overall experience of using cljdoc and give me more insights into what we thought we were handling but aren't!