How has this been looking for everyone?
The automatic bug reports have dropped significantly - the only thing that's come it is from one user, and only when using the mark as correct button, so it *seems* (crosses fingers) to be much better.
From your end(s), how has it been?
Funny you should ask---I just noticed this:
Today I only studied my two smaller vocab schedules (JLPT and Extras) and not my big Everything schedule. I assume that since I didn't study the Everything schedule it was never auto-flushed.
BN: You are indeed the only person left to be sending me automated reports at this point. Was going to take a look at it tonight.
shvelven - if I understand it correctly, Everything = JLPT + Extras? So any change in the smaller should create an instant change in the larger as well?
If you flush the Everything, do both numbers adjust, or does one of them stick to what it is showing?
PS: That is the last thing I'd call funny.
You're correct. My JLTP shcedule has JLPT N5 and N4, my Extras schedule has all of my private lessons, and Everything has N5, N4, and all of my private lessons. I've been using the Everything schedule for two purposes: first is to get a quick look at my stats between all schedules (how many total words do I know, how many total questions are left for today, etc) and the other is to regulate new terms---I have the Everything schedule set to give me no new terms, so if I feel I'm falling behind I just study that schedule. Since I haven't been falling behind, I've been studying my two smaller schedules without touching "Everything".
So yes, studying either of my two smaller schedules should change the larger schedule.
There is some overlap between JLPT and Extras; that is, some of the terms in my private lessons are also on the JLPT. I'm not sure to what extent that would affect the remaining counts.
Haven't been seeing any bug reports today, but my schedule numbers (terms ready to study and actually studied terms) don't match.
Today 5233 gave me a 51 question quiz (it's set to 50).
Four others have been weird as well:
5078 Mon. 16rts 19studied. Tue. 22rts 23studied Wed. 24rts 26studied.
5702 Mon. 27rts 32studied. Tue. 25rts 27studied Wed. 28rts 32studied.
7368 Mon. 29rts 31studied. Tue--fine Wed. 27rts 28studied.
7369 Mon/Tue fine Wed. 13rts 16studied.
This discrepancy happens no matter where I start a quiz (dashboard, summary, dropdown).
You'll see a bug report from me today, but that's from marking "redo" on a kanji question.
I haven't been getting the error message, but I'm still getting a mismatch in my "Everything" schedule:
Just like last time, I've been studying my two smaller schedules without touching "Everything". Maybe you should put me back in the verbose logging group?
I've been away for two days, but have gotten a ton of notices in my email unfortunately. I'm going to look over them. However, if I cannot find the info I need, I'm going to switch over from high level (marking points in the code) to low-level (logging database updates on the relevant tables), as I'm nearing the end of my so-called rope.
Even though it's been a week since the last update, I've been working feverishly behind the scenes. I've been continually adding more logging, and have just set up an additional fine-grained layer to help track what's going on.
I've also been tracking just how much processing time the actual schedule updates are taking, and even when a user hits a term that updates across multiple schedules, it seems to be quite speedy.
I might try simply moving that processing back 'internally' inside of the quiz - there are always potential issues when you take a process, serialize part of the data, and save it to be unserialized and processed on a different server, so this might be of use. Of course, I am not interested in doing so if it creates a noticeable delay in the quizzing speed.
Until then, though, I've noticed a few more things that are running slow in the back that would normally not affect you guys, but are affecting those that are being logged due to the way the logging has to be run, so I'll try to clean that up so the logging is not as slow as it has been.
BN: No problems at all!
I have some exciting news - i discovered 2 key points where I found solid evidence of problems that were causing issues.
One was a brand new issue I had not considered until now - I had basically gotten a bit too optimistic about future optimizations and had implemented something on the backend for a future update without considering a few edge cases that render it useless. That is unfortunate, but I'd rather have a slightly slower backend than one that doesn't work.
The second involves what I mentioned above - taking the processing and not 'outsourcing' it to a back server, instead handling it within the quiz itself. There were some communication issues taking place that were causing things to be updated twice.
I've put all the new changes in, and the loggers are still active, so now I need to wait and see how much of a change this makes.
Still, I feel optimistic!
I've made some AWESOME catches using my new logging system - I finally found a level of logging that works extremely well in showing me exactly where the problem is. I've nailed down a few edge cases, and I've seen the number of reports plummet.
If any of you are a)not having me log you and b)still having significant issues, please let me know so I can (temporarily, i promise!) throw you into the logger.
For those of you in the logger and if/when you get an issue - please let me know if you had a question 'freeze' (read: you had to submit the answer twice) in the previous question or perhaps once before that.
With a few other patches I've put in, it looks like (hopefully) the only errors being caused now are actually coming from the debugging taking too long and stalling the quiz in rare cases. If I don't get any 'new' types of errors, I'm going to turn off the logging in a day or so and see how it runs.