why did sympy-bot add double entries for some PRs in the release notes?

[dandiez]

https://github.com/sympy/sympy/wiki/Release-Notes-for-1.4/_history
There seem to be double entries for sympy/sympy#16063, sympy/sympy#15816, sympy/sympy#16028...

[asmeurer]

Thanks for pointing this out. Please open bot issues in this repo in the future.

[dandiez]

ok, will do!

[asmeurer]

I can't tell what is happening here. It seems to be getting the "closed" webhook twice. I'll need to merge #50 to debug it (otherwise finding specific webhooks on the GitHub webhooks page is a nightmare).

[asmeurer]

Here's the exact same webhook being received twice:

2019-02-25T22:40:13.203297+00:00 app[web.1]: Received pull_request event with delivery_id=4a5c8340-394e-11e9-9998-280fe36e9124
2019-02-25T22:40:13.204205+00:00 app[web.1]: PR #16071 was synchronize.
2019-02-25T22:40:13.438722+00:00 app[web.1]: Received pull_request event with delivery_id=4a5c8340-394e-11e9-9998-280fe36e9124
2019-02-25T22:40:13.438879+00:00 app[web.1]: PR #16071 was synchronize.
2019-02-25T22:40:14.078943+00:00 heroku[router]: at=info method=POST path="/" host=sympy-bot.herokuapp.com request_id=5b3976bc-d390-4bf7-9cd5-504ab2290f53 fwd="192.30.252.35" dyno=web.1 connect=1ms service=879ms status=200 bytes=173 protocol=https
2019-02-25T22:40:14.282360+00:00 heroku[router]: at=info method=POST path="/" host=sympy-bot.herokuapp.com request_id=84e7e904-5fc3-49b7-859e-ccbfb56a40e2 fwd="192.30.252.35" dyno=web.1 connect=1ms service=848ms status=200 bytes=173 protocol=https

[asmeurer]

I don't know if this is a coincidence, but every duplicate webhook that I have found was due to an action by @smichr. I did find some @smichr actions that weren't duplicated, like closing sympy/sympy#16071 and merging sympy/sympy#16004. I have no idea why this would be the case if it isn't a coincidence.

[asmeurer]

I pushed a change the the wiki removing the duplicate entries. I haven't seen this happen again, either for merge or synchronize events. I'll keep my eye on it, but perhaps (hopefully) it was just a fluke with the GitHub API.