[07:35:32] The problem is in Z19679, which is invisibly broken 😱 (re @u99of9: I'm heading to bed and unfortunately I've left Z27481 hanging. I don't know if it's my error or another system issue.) [07:36:56] https://tools-static.wmflabs.org/bridgebot/75465e21/file_73142.jpg [07:37:55] oh wow - ghost implementation responsible? : https://tools-static.wmflabs.org/bridgebot/cd1f6f85/file_73143.jpg [07:39:24] That would do it… but why is it the first listed? 🤷‍♂️ [07:40:29] It was deleted last year!! https://www.wikifunctions.org/wiki/Z19755?uselang=en&action=history [07:41:47] because you disconnected the other two in January, leaving it at the top, before the others were reconnected! https://www.wikifunctions.org/w/index.php?title=Z19679&diff=160535&oldid=160530 (re @Al: That would do it… but why is it the first listed? 🤷‍♂️) [07:42:35] no idea why it only stopped working now though [07:46:39] Yeah, tests were successful on July 6th 🤷‍♂️ [07:55:27] I've reopened *T379873* for starters. [07:58:29] How did you find this? (re @Al: The problem is in Z19679, which is invisibly broken 😱) [08:06:08] I edited the implementation, simplifying each leg until it gave incorrect results rather than an error. Then I cancelled that and edited again focusing only on the faulty leg… (re @u99of9: How did you find this?) [08:11:20] Sorry to make you do all that! (re @Al: I edited the implementation, simplifying each leg until it gave incorrect results rather than an error. Then I cancelled that an...) [08:13:19] Not your fault. I’d’ve found the problem sooner if I hadn’t assumed “add rational numbers” was working 😏 (re @u99of9: Sorry to make you do all that!) [08:15:23] Yes that's not the most obvious suspect, especially with all green ticks. Because the error was something like "no connected implementations" I had gone through every function the composition calls, but had seen lovely green everywhere. (re @Al: Not your fault. I’d’ve found the problem sooner if I hadn’t assumed “add rational numbers” was working 😏) [08:27:48] Yes, I don’t know why it doesn’t say for which function. Or why successful implementations don’t get promoted above failing ones by WikiLambda system. [09:16:23] Now that it's documented, I think I should manually fix this function. It's just too important to sit around broken. [09:17:38] I kinda fixed it already. It now works even though the ghost is still there. (re @u99of9: Now that it's documented, I think I should manually fix this function. It's just too important to sit around broken.) [09:53:23] Looking at it, I’m not sure the logic is correct. The rational number argument represents a date in the proleptic Julian calendar. Conversion to proleptic Gregorian would require a variable adjustment according to the number of leap days omitted in the Gregorian calendar (eleven from 1600 to 2399, for example). (re @u99of9: I'm heading to bed and unfortunately I've left [09:53:23] Z27481 [09:53:23] hanging. I don't know if it's my error or another system issue.) [09:57:10] Try the converter at https://aa.usno.navy.mil/data/JulianDate (re @Al: Looking at it, I’m not sure the logic is correct. The rational number argument represents a date in the proleptic Julian calenda...) [09:59:03] I think you've overcomplicated it. It's just a sequence of days, and Z20750 should handle all the complexity. (re @Al: Looking at it, I’m not sure the logic is correct. The rational number argument represents a date in the proleptic Julian calenda...) [10:00:35] In fact even my implementation overcomplicates it by splitting around the transition (because at the time I wasn't thinking proleptically...). I could have just started at JD=0 and added Q days. [10:03:50] Yes. But then you’d end up with a Julian calendar date, not a Gregorian calendar date. The divergence has to be accounted for somewhere, just not with a constant. (re @u99of9: In fact even my implementation overcomplicates it by splitting around the transition (because at the time I wasn't thinking prol...) [10:10:20] I would choose the Gregorian date of JD=0 as my starting point. Then Z20750 is a Gregorian move, so should always get the right sequence of dates up to infinity. (re @Al: Yes. But then you’d end up with a Julian calendar date, not a Gregorian calendar date. The divergence has to be accounted for so...) [10:10:46] Can you propose some tests that you think it will get wrong? [10:12:24] (2597701 converts to “A.D. 2400 February 29” which is not a Gregorian date.) (re @u99of9: Try the converter at https://aa.usno.navy.mil/data/JulianDate) [10:18:22] but it is because it's divisible by 400? [10:19:58] Exactly. Presumably it’s March 12th in the Gregorian calendar. (re @u99of9: but it is because it's divisible by 400?) [10:25:29] No, I think the site *is* giving Gregorian calendar dates (as is my function). Here are some tests to demonstrate. Z27488 and Z27489 have the input JD value incremented by 1. But if the output was Julian, the second would have a 29 February. [10:25:54] https://aa.usno.navy.mil/calculated/calendardate?ID=AA&jd=2488128.02252&submit=Get+Date [10:25:58] https://aa.usno.navy.mil/calculated/calendardate?ID=AA&jd=2488129.02252&submit=Get+Date [10:26:11] Well, it’s wrong for 2597701, then. (re @u99of9: No, I think the site *is* giving Gregorian calendar dates (as is my function). Here are some tests to demonstrate. Z27488 and Z2...) [10:27:06] Why? Gregorian years divisible by 400 get a leap day back. So 29 Feb is correct. (re @Al: Well, it’s wrong for 2597701, then.) [10:27:38] Oh, never mind… that’s just my head! (re @Al: Well, it’s wrong for 2597701, then.) [10:29:02] overcomplicating! [10:31:37] Maybe, but there is still divergence prior to 1582, where the site uses Julian calendar dates. (re @u99of9: overcomplicating!) [10:43:51] Ah, yes, I hadn't looked into that yet. (re @Al: Maybe, but there is still divergence prior to 1582, where the site uses Julian calendar dates.) [12:01:00] Unfortunately it didn't somehow also fix Z25356 (re @Al: I kinda fixed it already. It now works even though the ghost is still there.) [12:05:15] I think the composition is okay when run 🤷‍♂️ (re @u99of9: Unfortunately it didn't somehow also fix Z25356) [12:11:25] Oddly, Z27449 failed in edit but passed after Publish 🤷‍♂️ (re @u99of9: Unfortunately it didn't somehow also fix Z25356) [12:17:31] ha, just as I was writing up T402079. Oh well, I'll leave it there for a bit while the cache oddities resolve themselves?! (re @Al: Oddly, Z27449 failed in edit but passed after Publish 🤷‍♂️) [12:22:30] Yeah, just re-title it “All gone a bit weird” 😏 (re @u99of9: ha, just as I was writing up T402079. Oh well, I'll leave it there for a bit while the cache oddities resolve themselves?!) [12:25:56] I can’t help thinking those Z503s and Z504s are connected with “general stubbornness” 🤔 (re @u99of9: at Z10000 I'm getting a Z503 error!)