« first day (1181 days earlier)      last day (2638 days later) » 

10:09 PM
I'm taking Smokey temporarily for a test.
(in a few minutes)
 
!!/alive
 
@JanDvorak Of course
 
FR: standby instances should reply to !!/alive as well (active/standby and location)
 
@JanDvorak technically difficult; standby instances don't have chat connections and probably shouldn't
 
How are they awoken then?
 
10:13 PM
If SE chat goes down and that causes us to error, we don't need all the standby instances to also die.
@JanDvorak metasmoke tells them to.
All they do is send pings every minute. If needed, metasmoke will respond with a failover code
 
[ SmokeDetector ] SmokeDetector started in reverted mode at rev eeaaad4 (Undo1: Sort arrays in debug printing) (running on Undo's Raspberry Pi)
 
!!/standby teward
 
teward/Aurora is switching to standby
Restart: API quota is 8569.
 
...
!!/standby teward
 
teward/Aurora is switching to standby
 
10:16 PM
hrmph
still some kinks to iron out there
 
Maybe MS could list the instances through the active one?
 
Restart: API quota is 8568.
 
As in, "4 instances active. Foo/bar is active, foo/baz, quux/quux and xyzyx/xyz are on standby"
 
You could, but that's complexity
 
hm let me check something
 
10:20 PM
[ SmokeDetector ] SmokeDetector started in reverted mode at rev eeaaad4 (Undo1: Sort arrays in debug printing) (running on Undo's Raspberry Pi)
 
Whoa, that's a nice commit hash
 
@Undo there may be a logic error on nocrash.py
 
@JanDvorak Oddly satisfying, yeah
@ThomasWard I've thought there was, but I haven't been able to find it upon looking
 
@Undo it looks like it failed and did first_start
@Undo livetracking htop seemed to help a little
 
@ThomasWard what failed, though?
 
10:23 PM
!!/location
 
@ThomasWard teward/Aurora
@ThomasWard Undo's Raspberry Pi
[ SmokeDetector ] SmokeDetector started in reverted mode at rev cdc7098 (Undo1: Fix KeyError) (running on Undo's Raspberry Pi)
 
@Undo i'm writing in debug code at the moment to see what happens
 
Restart: API quota is 9937.
 
with a printout of some messages, a true/false, etc.
@Undo I think i fixed the logic error but we'll see
 
@ThomasWard Feel free to commit, I'm not touching nocrash anyway so there won't be any merge issues
 
10:27 PM
[ SmokeDetector | MS ] Repeating characters in answer: How do I collect loot stored in my town hall in COC? by Mike on gaming.stackexchange.com
 
@Undo already pushed, waiting for it to come back
with a CI Success
 
CI on a6add58 succeeded.
CI on a6add58 succeeded.
 
!!/standby undo
 
Looks like it worked
 
[ SmokeDetector ] SmokeDetector started in reverted mode at rev cdc7098 (Undo1: Fix KeyError) (running on Undo's Raspberry Pi)
 
10:27 PM
uhm
 
... nope
 
i think i broke mine
 
manually killed mine
 
just killed Aurora with a stick now
 
10:28 PM
[ SmokeDetector ] SmokeDetector started at rev a6add58 (Thomas Ward: nocrash.py: Logic error fix, debug code) (running on teward/Aurora)
Restart: API quota is 8520.
 
!!/standby teward
 
Worked, looks like
 
well I guess I did maybe fix it
let me check one thing
 
10:29 PM
oops I broke KeyboardInterrupt
 
@SmokeDetector naa was strangely appropriate for an answer containing multiple 💩💩💩 characters :)
 
I've just been killing it in htop, it's easy anyway
 
ArtOfCode/EC2 received failover signal.
Restart: API quota is 9999.
 
Well that part works
!!/rev
 
10:31 PM
@Undo i'm going to launch locally, location name = teward/Overlord for a few moments
 
kk
 
because I need to not track shit over SSH :P
 
CI on efaa17c succeeded.
 
!!/pull
 
[ SmokeDetector ] SmokeDetector started in standby mode at rev efaa17c (Thomas Ward: More logic fixes) (running on ArtOfCode/EC2)
 
10:32 PM
@SmokeDetector that's broken
 
@Undo because Art never moved to the new codebases
for nocrash
 
oh, right
@ArtOfCode at some point, need manual pull to get nocrash fixes
 
ArtOfCode/EC2 received failover signal.
 
urgent?
 
nah
 
10:33 PM
Restart: API quota is 9990.
 
k, it can wait then
 
!!/rev
 
!!/standby art
 
ArtOfCode/EC2 is switching to standby
[ SmokeDetector ] SmokeDetector started in standby mode at rev efaa17c (Thomas Ward: More logic fixes) (running on ArtOfCode/EC2)
Restart: API quota is 9999.
 
10:34 PM
alright, first...
!!/standby teward
 
[ SmokeDetector ] SmokeDetector started in standby mode at rev efaa17c (Thomas Ward: More logic fixes) (running on teward/Overlord)
 
OK so that's why that happened
 
afk for a while. Superping if you need something really bad @ThomasWard.
my Pi is down too
 
ArtOfCode/EC2 received failover signal.
[ SmokeDetector ] SmokeDetector started at rev efaa17c (Thomas Ward: More logic fixes) (running on teward/Overlord)
 
10:35 PM
!!/standby art
 
ArtOfCode/EC2 is switching to standby
[ SmokeDetector ] SmokeDetector started in standby mode at rev efaa17c (Thomas Ward: More logic fixes) (running on ArtOfCode/EC2)
Restart: API quota is 9998.
CI on e5caa59 succeeded.
 
!!/pull
 
Restart: API quota is 9992.
Restart: API quota is 9991.
Restart: API quota is 9988.
 
Did you break it?
 
@ByteCommander fixing a few errors in logic handling in nocrash.py
that's me ctrl+c-ing and restarting it a lot
 
10:42 PM
!!/blame
 
Restart: API quota is 9985.
 
!!/location
 
@ThomasWard teward/Overlord
 
Might be better to use the signal module instead of just catching KeyboardInterrupt. More flexible.
 
@ByteCommander ctrl+c triggers KeyboardInterrupt
all other exceptions raise as they should
 
10:44 PM
But you can intercept the SIGINT or other signals by setting a signal hook, so that it doesn't throw a KeyboardInterrupt at all.
 
[ SmokeDetector ] SmokeDetector started at rev 68103af (Thomas Ward: Keyboard Interrupt handling, round 2) (running on teward/Overlord)
 
@ByteCommander you want to make a pull req for that? I'm lazy :P
 
Restart: API quota is 9983.
 
it also needs to be handled in the try/catch that surrounds the subprocess
 
If you tell me what exactly you want to achieve... maybe.
But probably not tonight.
 
10:45 PM
well, it kinda works and just dies off which is fine
 
[ SmokeDetector ] SmokeDetector started at rev 68103af (Thomas Ward: Keyboard Interrupt handling, round 2) (running on teward/Overlord)
 
@ByteCommander buzzer
try again, we're Py2
 
Phh... Upgrade it.
 
some things don't behave in py3 yet
 
10:48 PM
Same module, but uglier doc page: docs.python.org/2.7/library/signal.html
 
[ SmokeDetector ] SmokeDetector started at rev 32ea01f (Thomas Ward: Remove debugging print outputs.) (running on teward/Aurora)
 
well, i think that fixes it.
 
Restart: API quota is 8519.
 
!!/standby teward
 
[ SmokeDetector ] SmokeDetector started in standby mode at rev 32ea01f (Thomas Ward: Remove debugging print outputs.) (running on teward/Aurora)
[ SmokeDetector ] SmokeDetector started at rev 32ea01f (Thomas Ward: Remove debugging print outputs.) (running on teward/Aurora)
 
10:49 PM
yay
 
Restart: API quota is 8518.
 
@Undo for the record: fixed the logic check for whether we run in standby or not, and also poked some things to make keyboard interrupts like ctrl+c work right... sorta.
i might have to write it in ws.py actually to have a try/except around most of the code to capture KeyboardInterrupt and exit with a specific error code that nocrash.py can interpret better
but standby/active works again.
 
@ByteCommander [status-planned-but-probably-not-going-to-happen-cause-its-a-big-job]
 
Are you sure? Getting ChatExchange 2&3-compatible wasn't such a big deal either...
I guess it's more a question of mentality.
 
@ByteCommander i wasn't talking about CE in this case
but if you want to talk about CE, how about we figure why the debugger can't get the login cookie when it's being run in a debugger.
 
10:56 PM
[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title: What IDE options do we have, to choose from, when we are working in a software company in India? by Zakstar on stackoverflow.com (@dorukayhan)
 
blame the debugger...
 
i'll blame someone else
!!/blame
 
@ThomasWard It's Byte Commander's fault.
 
oh lookit
 
Can it be that the debugger breaks on exceptions, even if they would be caught later?
 
10:57 PM
@ByteCommander nope, because then it'd 'suspend' and wait for me to do something
so yeah evil
ANYWAYS
i should probably head home, it's dark, and approaching the time of Reckoning
 
[ SmokeDetector | MS ] Pattern-matching website in answer: Embed Dropbox website to an HTML website by Rob on stackoverflow.com (@dorukayhan)
 
J F
o/
 
@SmokeDetector Oh dear. I autoflagged this as spam. Its a FP :/
I have retracted my (auto)flag ...
 
Are you sure it's not a seed?
 
@JanDvorak Doesn't look like it.
@SmokeDetector why
 
11:12 PM
Title - Position 80-96: company in India
Body - Position 32-48: company in India
 
that's... kinda what happens when you have 180 as a weight
it's probably still fairly accurate, but more FP's than something higher
@SmokeDetector f
 
My strategy to not opt into autoflagging seems to be working out so far
 
@ArtOfCode 180 is 99.73%.
 
so there's one of your 27 false positives per 10k posts
 
J F
Speaking of autoflagging, could someone set me up for it?
 
11:21 PM
Proposal: bump the required accuracy up to 99.5%
 
Changed to 240 (99.92%)
 
@JF have you got a MS account?
 
@angussidney ? That is less than what I had it set to.
 
J F
@angussidney Yes. ID 65.
 
@DavidPostill currently the required accuracy is 99%, which is a little to inaccurate IMO
@JF 2secs gotta give you the right permission
Okay, now go to the flagging button in the toolbar that should appear if it hasn't already
 
11:25 PM
the flagging link is always there
 
Next click the autoflagging conditions link and create a new condition
 
there's a guide through the whole thing in the wiki
 
(recommended weight is 280, which gives you 99.98% accuracy)
@ArtOfCode can you get a link to that? It might be a little hard to get to on mobile
 
@angussidney Probably should bump it up, yeah
 
11:28 PM
@Undo done
 
[ SmokeDetector ] SmokeDetector started in standby mode at rev 32ea01f (Thomas Ward: Remove debugging print outputs.) (running on Undo's Raspberry Pi)
 
Now, we have an issue about existing conditions
Best option is probably to disable them if they don't meet the new condition
 
Is this correct?
"There's a threshold of accuracy that all flagging conditions must meet before they're used to cast flags automatically. Currently, this is 99.9%"
 
@DavidPostill no
Probably shouldn't have a dynamic number in a static document
 
J F
Thanks all!
 
11:30 PM
Also @JF can you let us know whether or not the sitepicker appears when you go and set up preferences and conditions? We had a bug where it wasn't showing up, and were not sure if we have fixed it
If it doesn't show up, you can refresh the page and it should
 
J F
@angussidney I already set up preferences a while ago. It does work for conditions now.
Do I have to fill in all of the fields in conditions?
 
yep
280/1/1 is recommended
 
J F
11:56 PM
[feature-request] make the fields default to 280/1/1
 
@JF probably a good idea
 

« first day (1181 days earlier)      last day (2638 days later) »