Some fights pages return an internal error about antlog HOT 14 CLOSED

from antlog.

This issue is caused by missing records in the entrants table.
Only AWS53 and AWS54 are affected.
Entrants 497 (group 6), plus 442 and 468 (group 7) are missing from AWS53 (event 7)
Entrant 605 is missing from AWS54 (event 9)
There are other entrants "missing" but this is OK because they don't appear in the fights table, so they have been deleted, probably fleas or nanos that were signed up for ant events then withdrawn.

from antlog.

Found the root cause.

Some of the entrants are being exported from the local version of Antlog with the status field set to -1 even though they have taken part in the event and should end up with status = 0, 1 or 2.
At the end of the SQL file all entrants to complete events are deleted if their status is -1 (did not compete).

So the question is, why does the status end up at -1 for entrants that have competed?
This only seems to have happened since AWS53 (November 2017)

from antlog.

Something relating to Chris' virtual machine install perhaps? As always, I have no idea what I'm talking about but try to offer vague suggestions ;)

from antlog.

I'd hesitate to suggest it's Chris's fault, there were probably changes made to Antlog between AWS50 when it was last used without issue and AWS53 when the problem first appeared.
I've now fixed the Antlog database, I managed to recover the missing records from old copies of the database.

from antlog.

Aha! That has actually proved related to the reason I was looking into the fight list...
Entrant 605 is The violation of our most basic human rights

Group 7 Round 3, Loser's Bracket, Pingu beats TVOOMBHR.
However, strangely, in Round 4 TVOOMBHR was drawn instead of Pingu.
That Round 4 fight didn't actually happen as the team had rightly assumed all their robots were out and had left. Seems some fault had put them through and Pingu out despite Pingu's win.

@cjsoftuk

from antlog.

Yes I noticed that, and I don't ywt know why it happened, but it might explain why the entrant record was exported with status = -1. The status is used to track the "lives" in double elimination. It starts at 2 and decrements when you lose. Everyone should end up at zero except the winner. Since TVOOMBHR lost 3 fights its status would get to -1.
However this doesn't explain what happened at AWS53 as 3 entrants ended up with status = -1 but all of their fights & results look correct.

from antlog.

@GaryA - Any news on this one? This one's probably the most crucial to be fixed for AWS55.

from antlog.

@cjsoftuk I'm struggling to find and reproduce the root cause of the error. I assume it was caused by a result being changed, but the entries in the log don't quite tie up with what normally happens and when I reply the queries in the log I still don't get to the erroneous state.
I will try recreating the event and logging all the results up to where the wrong entry was changed, but it's a lot of boring work.
Every time I've tested changing a result it has worked perfectly. It's a necessary feature as there will always be finger trouble, and I can't find any issue with the logic so far.

from antlog.

Thanks Gary - I agree that it's boring to go through that - am happy to do the testing there if needed. I'm intending to spend all of next weekend doing prep for AWS55 so if we don't have any progress by then I will try and do some more digging.

from antlog.

from antlog.

Rebuilt database as it was at the start of AWS54, with the draw done but no fights run.
Ran all fights, including Group 7 Round 3, Loser's Bracket, Pingu beats TVOOMBHR, but recorded as TVOOMBHR winning.
Ran a couple more fights then went back and corrected Group 7 Round 3 Loser's Bracket to set Pingu as winner.
Continued with fights and Group 7 Round 4 was correct.
At the end all the data was correct, all entries had status = 0 except winner (status = 2). No entries with status = -1

Don't know how to reproduce the problem....

from antlog.

Very confused how that happened then (doubly odd since it didn't happen at AWS53 and there wasn't a lot between them code-wise). I guess we'll just have to keep a close eye on it to see if it happens again.

from antlog.

from antlog.