I upgraded to Apollo 3.7 and set up used coordinates, and ran into something which might be a bug, b...

# apollo-kotlin

[Eduard Boloș]

I upgraded to Apollo 3.7 and set up used coordinates, and ran into something which might be a bug, but I am not sure. Either way, it was interesting. Basically, given an interface A { foo: X, bar: Y} and a type that implements it type B implements A { foo: X, baz: Z}, and a fragment F on A { foo, bar, ...on B { baz }}, then BBuilder will only have a field for baz . Shouldn't foo and bar be also added to the builder? How else would you define their values? 😄

[Eduard Boloș]

Btw, my workaround was to re-write the fragment like fragment F on A { ...on B { foo, bar, baz }}

[mbonnin]

That makes sense.

[mbonnin]

Another workaround is to add the B.foo and `B.bar`coordinates to alwaysGenerateTypesMatching

[Eduard Boloș]

oh, yeah. but I guess that the expected thing would be to have them added to the builder by default, no?

[mbonnin]

yea, you're 100% right

[mbonnin]

every interface field needs to pull all the implementation fields

[mbonnin]

Looking into it

[Lukáš Kúšik]

I think we encountered this too..

[mbonnin]

We do have some code though...

[mbonnin]

Mmm this is done while merging the fields but not when computing used types from other modules

[mbonnin]

https://github.com/apollographql/apollo-kotlin/pull/4506/files

[mbonnin]

If that works for you, we'll make a patch release

[Eduard Boloș]

I can test it tomorrow, for now I went with your suggested workaround 🙂

[Eduard Boloș]

Hmm, I am trying to run publishToMavenLocal, but I get a bunch of errors:

* What went wrong:
Script compilation errors:

  Line 001: import JapiCmp.configureJapiCmp
                   ^ Unresolved reference: JapiCmp

  Line 094:     checkGitStatus()
                ^ Unresolved reference: checkGitStatus

  Line 119: rootProject.configureNode()
                        ^ Unresolved reference: configureNode

  Line 120: rootProject.configureJapiCmp()
                        ^ Unresolved reference: configureJapiCmp

Any idea what could be wrong? 😞

[Eduard Boloș]

It seems like Gradle doesn't pick up build-logic 😕

[mbonnin]

Mmm, that's unexpected...

[mbonnin]

What commit are you on?

[mbonnin]

(I'm at adccf1360de92d00e7ce93d4c5a825532f310769 )

[Eduard Boloș]

same: Your branch is up to date with 'origin/fix-used-coordinates-on-interfaces'.

[Eduard Boloș]

openjdk version "11.0.11" 2021-04-20
OpenJDK Runtime Environment AdoptOpenJDK-11.0.11+9 (build 11.0.11+9)
OpenJDK 64-Bit Server VM AdoptOpenJDK-11.0.11+9 (build 11.0.11+9, mixed mode)

Intel Mac, if it matters

[mbonnin]

rm -rf */build **/**/build

[mbonnin]

I've seen it a couple of times where the build-logic jar becomes empty

[mbonnin]

Not 100% sure what the cause may be

[Eduard Boloș]

well, that didn't help. build-logic.jar gets regenerated, but it's only 884 bytes, is that normal?

[Eduard Boloș]

unzipped it, it only has MANIFEST.MF inside it 🙈

[mbonnin]

--no-build-cache maybe?

[Eduard Boloș]

--rerun-tasks did the trick 🙂 thanks!
I will let it build, and try it out after lunch

[Eduard Boloș]

Ok, can confirm, the fix works well 🙂

[mbonnin]

Nice, thanks for confirming! Will do a patch on Monday