Fork me on GitHub
#clojars
<
2017-11-01
>
juhoteperi16:11:33

Any idea what happened with this package: http://repo.clojars.org/cljsjs/react-select/ https://clojars.org/cljsjs/react-select ? rc.10 is uploaded to the repo, but it is not shown on web Deploy threw this error:

Could not transfer metadata cljsjs:react-select/maven-metadata.xml from/to clojars (): Access denied to: , ReasonPhrase: Forbidden - Serialized task failed.

tcrawley16:11:35

@juhoteperi Interesting, I'm not sure what that's about. The maven-metadata.xml shows 1.0.0-rc.10, do you happen to still have the full output of the deploy attempt?

tcrawley16:11:33

I can see the output, thanks

tcrawley16:11:59

I don't see any errors in clojars' log, nor in sentry, so it's unclear why this failed

tcrawley16:11:40

The "Serialized task failed" makes me think it had a problem writing to the db, possibly

juhoteperi16:11:40

Is that message from Clojars backend?

tcrawley16:11:51

to support transactional deploys, we use the writing of the metadata file as a trigger to copy the tmp repo where we store your upload over to the real repo

tcrawley16:11:59

and do the validation/db updates at that point

tcrawley16:11:11

so this failure is likely unrelated to the metadata file itself

tcrawley16:11:12

It looks like the release is there, but not listed in the db, so it doesn't show in the UI: https://clojars.org/repo/cljsjs/react-select/1.0.0-rc.10/

tcrawley16:11:46

Would you be willing to try releasing an rc11 to see if this was a temporary issue?

juhoteperi16:11:25

building 1.0.0-rc.10-1 now (rc10 is the version of upstream package, don't want to touch that)

tcrawley16:11:35

ah, understood

juhoteperi16:11:05

Strange that the logs wouldn't contain this exception

juhoteperi16:11:48

repo routes seem to correctly catch-and-rethrow the exception, and repo routes are wrapped with repo/wrap-exceptions which should report them

tcrawley16:11:22

Glad it worked. Yeah, I agree - that should have gotten logged to stdout and to sentry, but it didn't go to either

juhoteperi16:11:01

aha... I see it

juhoteperi16:11:24

Maybe that is to skip errors caused by e.g. trying to deploy same version again

tcrawley16:11:39

good catch, we shouldn't be doing that for all errors, the intent there was to not report validation errors to sentry, since those are user errors

tcrawley16:11:11

would you mind filing an issue for that?

juhoteperi16:11:33

Yeah I'll open one