@email@example.com (#horh62a) Maybe at the weekend, but I’m rather busy these days with birthdays coming up (and/or no energy left after work to debug software 😳). But don’t worry, we’ll get this sorted out. 😊
@firstname.lastname@example.org (#pe5ymoa) Alright, I can’t tell when I’ll be able to do a screen sharing thingy. So let’s try this the old-fashioned way first. Please try to reproduce the issue with the branch quark-trace that I pushed recently. It’ll create a /tmp/jenny.log file (it will get large). When you see duplicate twts, try to find them in that log.
Reasons, in theory, why we could see dups:
1) jenny doesn’t detect your feed’s URL correctly.
2) python-dateutil doesn’t parse your twt’s timestamp correctly. Or rather, it parses it differently depending on some env vars? Cronjobs often have this pitfall where some env var is different than your normal environment.
Actually … I can’t think of anything else. 🤔 You don’t see dups all the time, it only happens for your own twts, and you said that the twt hash mismatches. That already narrows it down to something in make_twt_hash(). 🤔
Let’s see if that trace file helps. If it doesn’t, we can add more trace() calls.
@email@example.com (#pe5ymoa) Maybe her -00:00 timezones throws off a parser? I think I tried this edge case when coming up with the twt hash spec, but can’t tell anymore for sure. At least I’ve written this (2020-12-13T07:45:23-00:00 → 2020-12-13T07:45:23Z) explicitly down. The thing is, I just documented Go’s behavior to make sure I don’t break anything. Maybe add a test similar to this one and see what’s happening. :-?