https://stackoverflow.com/questions/44413952/gradle-implementation-vs-api-configuration suggests that findBugs should be implementation not api.

In D222#5368, @CLOVIS wrote:

In D222#5365, @Hackintosh5 wrote:

Findbugs should be compileonly.

That's what I thought but findbugs works on the generated JAR, if it's compileOnly findbugs cannot access it

You're right. But D217 is the way to go.

Findbugs should be compileonly.

I can add it probably, but I have to admit, I don't even know what coverage is!

This is not the way to do it. Compile is deprecated and should not be used.

In D218#5293, @CLOVIS wrote:

Is this ready for landing? Or do you plan to change other things

• Remove debug code and fix API running detection

• Don't kill the API if we didn't start it

In D218#5267, @CLOVIS wrote:

They sounds odd to me. I'll need to benchmark this on my machine to see exactly. You didn't forget to do the migrations, download the dependencies, etc?

In D214#5270, @CLOVIS wrote:

Oh. Okay, remove the "fixes t..." from the description here, otherwise pantocrator will close the task. I'm probably going to test all your diffs this week end.

In D214#5248, @CLOVIS wrote:

Looks good to me so far, but you need to do the Internal class too (in the same way).

Please tell me if you have a better name for it, "internal" doesn't record what it's for clearly enough in my opinion

In D213#5247, @CLOVIS wrote:

In D213#5219, @Hackintosh5 wrote:

@CLOVIS why have you rejected this?

I didn't so far, I just added myself as a reviewer. Are you sure this change works everywhere?

• Cleanup code

In D218#5234, @CLOVIS wrote:

In D218#5226, @Hackintosh5 wrote:

• Autostart the development API if it's not running already, and stop it again when we're done.

That's really not a good idea, the API takes 2-4 minutes to launch. Just abort if it doesn't run, that'll be much more convenient.

I've tried to land things before, it says I don't have access (403)...

• Remove debug and unused code

• Autostart the development API if it's not running already, and stop it again when we're done.

Push accidental code, now removed

Uff. No-one wants to parse a string JSON, how about an error code?

@CLOVIS why have you rejected this?

• Simplify companion object syntax

• Bring back the linters

In D216#5160, @CLOVIS wrote:

What are you even trying to do? The Nexus account is only used for deployment and you don't have the key anyway. It makes no sense to try to connect to an empty username and password

Undo accidental push, I'm still a noob

• Remove debug code

Fix arcanist being weird and pushing random unrelated code

• Typo

• Fix accidental upload of changed from another Diff

• Refactor WildFyre.kt to use a Java class and companion object

In D214#5119, @CLOVIS wrote:

In D214#5118, @Hackintosh5 wrote:

This is (from reading the Kt docs) the "correct" way... It's true, it would break existing clients except there is no-one using this library yet...

I don't like having global methods though, it's just not clean

This is (from reading the Kt docs) the "correct" way... It's true, it would break existing clients except there is no-one using this library yet...

Fixed locally, I'll submit a diff in the morning.

In D211#5079, @CLOVIS wrote:

In D211#5072, @Hackintosh5 wrote:

Um, idk how to run the tests. I *could* look it up if I weren't about to sleep. I'll look in the morning, but I assumed that it was simply a missing $, because that's what changed in D203.

That's weird. How did you manage to push this diff then? Arcanist doesn't normally let you diff when the tests do not pass. Please tell me what you did, there might a bug there.

Other than that, the procedure for the lib on Linux is:

• ./api.sh (let it run in the background)
• ./gradlew build
  
  On Windows:
• git submodule init
• git submodule update
• Run the API by following the steps on the wiki (or the contents of the script api.sh)
• gradlew.bat build
  
  Again, if you did not do this, there's a big problem because it shouldn't be possible to diff without doing those.

Um, idk how to run the tests. I *could* look it up if I weren't about to sleep. I'll look in the morning, but I assumed that it was simply a missing $, because that's what changed in D203.