Do you happen to know how https://github.com/felixdivo/ros2-easy-test/actions/runs/5035009110/jobs/9030212775#step:9:116 can happen in theory? i could not figure this out how even this happens...

I guess in my case it could be a race condition, since I access the executor from different threads, where I think it makes sense that it can occur. I accidentially found #1079, so maybe that is not supported anyways. It also does not happen determinisitcally. It just caused me to stare at these lines for a while and that's where I realized that there might be an issue.

EDIT: CC @fujitatomoya

Side note: Why are these signoff things required (I did read CONTRIBUTING.md)? I see basically no other projects use them and it makes contribting kind of a pain. E.g. we cannot add changes via comments and the defaults in most IDEs do not include this piece of text. Sure, no big deal but if I'm changing one line (e.g. some small typo, type annotation, ...) online and it has me clone the repo, undo commits, add the sign-offs, ... it makes contributing many times slower and often keeps me from fixing stuff because it is plainly annoying. Is this really legally required?

Do you happen to know how https://github.com/felixdivo/ros2-easy-test/actions/runs/5035009110/jobs/9030212775#step:9:116 can happen in theory? i could not figure this out how even this happens...

I guess in my case it could be a race condition, since I access the executor from different threads, where I think it makes sense that it can occur. I accidentially found #1079, so maybe that is not supported anyways. It also does not happen determinisitcally. It just caused me to stare at these lines for a while and that's where I realized that there might be an issue.

EDIT: CC @fujitatomoya

@felixdivo thanks for sharing. yeah probably that is not supported, #1079 is the same situation. but generating exception should be avoided even if it is not supported.

Side note: Why are these signoff things required (I did read CONTRIBUTING.md)?

In short, it is the lightest-weight thing we have to assure that contributors are allowed to contribute the code that they are providing. The other option would be a CLA, which is far more cumbersome and invasive.

E.g. we cannot add changes via comments and the defaults in most IDEs do not include this piece of text.

For what it is worth, you can always type it in by hand when making the commit message, and that will make the bot happy. Presumably you can also convince your IDE to append it to the commit message by default, particularly since it is a standard git commit flag (-s).

@felixdivo if you are comfortable with #1129 (comment), could you address DCO error? and i will start the CI.

@clalancette @fujitatomoya Thank you for explaining this. It's really a shame that we have to do this. I know you probably thought a lot about this, but a CLA is only signed once, which IMHO makes it easier. And regarding the legal consequences, it should not be more invasive.

Anyways, let's drop this discussion here since it has nothing to do with the changes in this PR. 😅

@clalancette are you good to go?

CI:

• Linux
• Linux-aarch64
• Windows