Update for 'messages-go v16' - WIP PR

[Yellaks]

I am able to get a successful build but 'go test' is failing on 8 test cases, 7 of which are failing due to:

• Error: runtime error: invalid memory address or nil pointer dereference

I am unsure on how to resolve these issues, any help and would be appreciated.

'go test' output attached: GoDog_test-results.txt

[Yellaks]

@aslakhellesoy - are you able to help me with this?

[Yellaks]

@aslakhellesoy - are you able to help here?

[Yellaks]

Open issue for this #396

[Yellaks]

@lonnblad - can you help in reviewing this pull request?

[mattwynne]

@cucumber/go is anyone able to help with this?

[mattwynne]

Thanks for your patience @Yellaks we'll get this figured out. I wish I could help but I'm totally new to Go.

[Yellaks]

@mattwynne - thank you

[mattwynne]

@Yellaks one option to unblock this would be for you to pair with me. I don't know Go but I do know the Cucumber ecosystem and perhaps together we could figure it out! You can book time with me at https://calendly.com/mattwynne or you can come and find us in https://cucumber.io/community#slack in the #committers channel to chat.

[eddie-knight]

Took a look at this today. Unfortunately I don't know a good way to debug the ...nil pointer dereference error, so I just went through all your changes to check for common issues I've seen cause that error (such as asking for X.y.Z when X.y doesn't exist). I made a few changes and ran the tests to see if I could find any fixes, but no luck yet on that front.

For reference, the issue is happening within this test: internal\formatters\fmt_output_test.go->Test_FmtOutput

...while parsing this feature file (but not others): internal\formatters\formatter-tests\features\scenario_outline.feature

[Yellaks]

@mattwynne @eddie-knight thanks for looking into this, I'm really keen to get this sorted, as I want to use GoDog but without the security violation. Is there anyone else we can ask to help look into this?

[mattwynne]

There's been some discussion around this issue in the community Slack, here.

Notably this from @aslakhellesoy:

@Eddie Knight I'm not sure where exactly you're getting errors, but this might be relevant: In the old protobuf implementation of messages, unassigned/empty primitive/array fields would always have a default value, even if you didn't set it. At least after deserialisation, but maybe also after direct initialisation of a message object (not entirely sure). With the new implementation, this isn't the case, at least not for the Go implementation (yet). So it could be that you're trying to read a message field that previously had a default value, but doesn't anymore. This could be the cause of the nil pointer dereferences you're seeing

[mbow]

@Yellaks I'm looking into why Argument: (*messages.PickleStepArgument)(), atm. will keep you updated.

[mattwynne]

Nice one @mbow!

Just a reminder that I'm happy to pair with anyone on this who needs a second brain on it. I've been part of working through this upgrade on the Ruby and JavaScript so I do have some context, even though I'm a novice Go dev!

https://calendly.com/mattwynne to book time with me, or ping me in Slack.

[mbow]

@Yellaks so have something working "ish" ( test suites pass ), but will pick it up next week, not far off just want to work out if these values were populated and with what values if they were not nil before, just because test suite passes but this could open another edge case for someone if step argument or doc has not been correctly set before.

[mbow]

@Yellaks please check and review https://github.com/cucumber/godog/pull/402

[Yellaks]

@mattwynne @mbow - Sorry I've not been online for awhile. Thanks both for looking into this, greatly appreciated. I have checked and reviewed #402, looks good to me :+1:

[mattwynne]

So with #402 merged can we close this one? @Yellaks we'd love to have you involved in the project so please feel free to find other itches to scratch, and submit a PR so we can get you the commit bit!

[Yellaks]

@mattwynne - thanks for closing this.