Evgeny Mandrikov

I guess that Java 10 will throw exception and use original interface definition in case if transformation adds method to interface and bytecode version < 8.

@marchof hope you don't mind that I also tried to do prototype - https://github.com/jacoco/jacoco/tree/issue-782 https://github.com/jacoco/jacoco/commit/775e17b20004ba5882112f8e546efe4cb944c1be :wink: It doesn't work in case of bad cycle with interfaces due to direct field...

> As for inlining, you could add documentation that it might improve performance to adjust the JVM's inlining thresholds. Idea is interesting. Unfortunately many people don't bother reading documentation and...

> The code uses a boolean array to store the coverage. Have you considered using static byte fields to store the coverage? Yes we considered: array-based probes up to 32767...

Thinking more about `$jacocoInit`: I guess that we already hurt inlining, because it calls our runtime. Thus wondering maybe better to sacrifize precision in case of bad cycle - move...

@chris2k2 according to https://github.com/jacoco/jacoco/pull/382#issuecomment-247153437 of @marchof this is just a proposal and work in progress: > proposed implementation -- **still WIP** As of today compared to other tasks and taking...

@marchof Let's don't forget that there is also Maven. BTW AFAIK Gradle integration can simply be seen as a wrapper over our Ant Tasks. In either case to me sounds...

As a user, between little inconvenience of "deep directory structure" and complicated configuration, I personally will prefer former option. As one of just few maintainers of JaCoCo, between accepting this...

@marchof do you have any thoughts about this ticket?

@marchof I can't find released version of agent with package `org.jacoco.agent.rt_mlrowg` , `rt_` prefix was used between 0.3.2 and 0.6.1 inclusive. So given the age of the ticket, absence of...