some crashes with missing virtual method

[chk2902]

Some customers of mine start to experience problems with the following stack trace:

STACK_TRACE=java.lang.NoSuchMethodError: No virtual method error(Ljava/lang/Exception;)Lcom/github/kittinunf/result/Result$Failure; in class Lcom/github/kittinunf/result/Result$Companion; or its super classes (declaration of 'com.github.kittinunf.result.Result$Companion' appears in base.apk!classes3.dex) at com.github.kittinunf.fuel.core.DeserializableKt.response(Deserializable.kt:171) at com.github.kittinunf.fuel.core.requests.DefaultRequest.responseObject(DefaultRequest.kt:490) at com.chk.go_elocal.database.flextariff.provider.Tibber.updateStatistics(Tibber.java:201)
caused as you can see by the call of the "request.responseObject" in Java code:

Triple<Request, Response, Result<FuelJson, FuelError>> result = request.responseObject(FuelJsonKt.jsonDeserializer()); Result<FuelJson, FuelError> responseResult = result.getThird();

Version is the current stable version 2.3.1, and it just happens in release builds (on Android).

implementation 'com.github.kittinunf.result:result:5.6.0' implementation "com.github.kittinunf.fuel:fuel:$fuel_version" implementation "com.github.kittinunf.fuel:fuel-android:$fuel_version" implementation "com.github.kittinunf.fuel:fuel-json:$fuel_version"
Strange: The same call works (in some other class) for the same customers when I use a full object path:

Triple<Request, Response, com.github.kittinunf.result.Result<FuelJson, FuelError>> result = request.responseObject(FuelJsonKt.jsonDeserializer()); com.github.kittinunf.result.Result<FuelJson, FuelError> responseResult = result.getThird();
although the "import" in the other class file has the same path.

Does anybody know what could be the reason? Ther was a gradle update at the end of november, but actually I don't know the build process good enough to know whether something like that could cause it...

[chk2902]

BTW: I did not change the Fuel version for a long time, so I guess it will be something in the build process.

Strange is: it works when it's called "manually" (called by a one-time Worker process), but (at least sometimes) it does not work in a scheduled Worker process.

Other idea: can it be that that "missing" method is called when there's a problem with the data passed to it, and I cannot reproduce it (and it has that strange pattern) as the data I get does not have that problem?