[16:15:06] Hello, all! There will be an announcement about this in the newsletter, but I wanted to drop a note here, too. We have been working on some major performance upgrades in the orchestrator and are planning to switch out to the new endpoint. The switch is slated to happen within the next 48 hours. Ideally, you shouldn't see any behavior changes except for 1) snappier function execution, 2) less flake/instability, and 3) [16:15:06] possibly some differences in errors. If you notice something bad, please file a task as a child of https://phabricator.wikimedia.org/T418887 . [22:35:27] There are some signs of bad stuff happening. The first I've seen is in the rational number reader. So if you go to Z19679 and enter strings, you sometimes get behaviour like this, seemingly when the reader is too slow for the UI and the UI just pushes on with weird representations. Sorry I can't report properly now. : [22:35:27] https://tools-static.wmflabs.org/bridgebot/a85a2c76/file_78537.jpg [22:35:28] https://tools-static.wmflabs.org/bridgebot/7b55d02a/file_78538.jpg [22:36:53] This is likely to do with the way we echo from helper functions, Z19949 (re @u99of9: There are some signs of bad stuff happening. The first I've seen is in the rational number reader. So if you go to Z19679 and en...) [22:45:34] Yes, good spot. Do other read functions follow a similar pattern? (re @u99of9: This is likely to do with the way we echo from helper functions, Z19949) [22:46:38] And errors everywhere in what was a mostly working function Z30000 [22:47:27] I tried the date readers at Z30000 which also failed. But since the actual well-specified tests failed too, there's likely another underlying problem. (re @Al: Yes, good spot. Do other read functions follow a similar pattern?) [23:05:21] I've been looking at Z30000, but as far as I can tell, the error affects the old endpoint, too 😞 . I will keep digging. [23:06:03] Please keep reporting here and/or in Phab, though; this is really helpful! [23:08:44] Yes, and when the read function fails, the expanded version contains function calls rather than literal objects. [23:08:44] I think this page of embedded functions was error-free yesterday https://www.wikifunctions.org/wiki/User:GrounderUK/Embedded_functions (re @u99of9: I tried the date readers at Z30000 which also failed. But since the actual well-specified tests failed too, there's likely anoth...) [23:14:22] Hmmmmmhmhmhm. So, what I see so far: [23:14:22] - if I test Z30000, I get failures in both versions, so I'm not sure what's going on there. [23:14:23] - when I run the embedded functions from that page in the UI, they succeed, so this is also very strange; I'll see if there's another reason this might be failing [23:18:07] Okay, `Z30000` errors are definitively unrelated (and actually point to a separate emergency :disap [23:18:12] Okay, `Z30000` errors are definitively unrelated (and actually point to a separate emergency 😞 . [23:18:23] I'm asking around about the embedded functions. [23:25:50] Hmmm… [23:25:50] Invalid programming language (0: "O", 1: "r", 2: "c", 3: "h", 4: "e", 5: "s", 6: "t", 7: "r", 8: "a", 9: "t", 10: "i", 11: "o", 12: "n", 13: " ", 14: "g", 15: "e", 16: "n", 17: "e", 18: "r", 19: "a", 20: "l", 21: "l", 22: "y", 23: " ", 24: "f", 25: "a", 26: "i", 27: "l", 28: "e", 29: "d", 30: ".") [23:25:52] From [23:25:53] https://www.wikifunctions.org/view/en/Z22475?call=%7B%22Z1K1%22%3A%22Z7%22%2C%22Z7K1%22%3A%22Z22475%22%2C%22Z22475K1%22%3A%7B%22Z1K1%22%3A%22Z39%22%2C%22Z39K1%22%3A%22Z1K2%22%7D%2C%22Z22475K2%22%3A%7B%22Z1K1%22%3A%22Z40%22%2C%22Z40K1%22%3A%22Z42%22%7D%7D (re @wmtelegram_bot: Okay, `Z30000` errors are definitively unrelated (and actually point to a separate [23:25:53] emergency 😞 .) [23:40:21] HAH! Okay, that one is actionable and definitely due to v2. Can you file a task with instructions? I hope to look at it tonight but need to run out for a bit. [23:44:51] I think I know what the issue is, but it's wild that it manifested in this way. [23:45:16] Yes. Doing it now. (re @wmtelegram_bot: HAH! Okay, that one is actionable and definitely due to v2. Can you file a task with instructions? I hope to look at it ...)