« first day (1217 days earlier)      last day (2603 days later) » 

12:00 AM
...what?
Anyways
@NobodyNada It is a little weird -- the user has legit answers and questions on other sites, but he's just joined today on a bunch of sites so it's odd
 
12:11 AM
@ArtOfCode do you have an example or a way to get that pickle? I won't get to it over the weekend, but I might be able to next week before vacation, and figure out something for that analysis
 
@Andy it doesn't exist, yet, it's not being run
 
OK. I'll wait then. :) I can look at it in a few hours
 
The basic analysis script I threw in just gives you min, max, and average for each site, so anything better would be good :)
 
That's a good start
 
@Andy actually, hang on. I can get at EC2, so I'll just pull in that branch and run it for a while
!!/reboot
 
12:15 AM
Goodbye, cruel world
[ SmokeDetector ] SmokeDetector started at rev 4d62414 (ArtOfCode-: Analysis script) (running on ArtOfCode/EC2)
Restart: API quota is 18605.
[ SmokeDetector ] SmokeDetector started at rev 95ee2fe (Thomas Ward: Changes to Post class, bodyfetcher.) (running on ArtOfCode/EC2)
 
that's got a bug
 
Restart: API quota is 18604.
[ SmokeDetector ] SmokeDetector started at rev c44e0bc (ArtOfCode-: Fix index out of range bug) (running on ArtOfCode/EC2)
[ SmokeDetector ] SmokeDetector started at rev 95ee2fe (Thomas Ward: Changes to Post class, bodyfetcher.) (running on ArtOfCode/EC2)
Restart: API quota is 18602.
 
that also has a bug
 
Bug, Bug, Bug, Goose
 
[ SmokeDetector ] SmokeDetector started at rev c44e0bc (ArtOfCode-: Fix index out of range bug) (running on Continuous Integration. If you're seeing this, create a config file with sane stuff in it.)
Restart: API quota is 19989.
 
12:26 AM
[ SmokeDetector ] SmokeDetector started at rev c44e0bc (ArtOfCode-: Fix index out of range bug) (running on ArtOfCode/EC2)
[ SmokeDetector ] SmokeDetector started at rev 95ee2fe (Thomas Ward: Changes to Post class, bodyfetcher.) (running on ArtOfCode/EC2)
 
bah, that's not fun
bug that I can't reproduce locally
 
!!/gitstatus
 
On branch master
Your branch is up-to-date with 'origin/master'.
nothing to commit, working directory clean
Restart: API quota is 18575.
 
@SmokeDetector on master
not deploy
Not that it should make much of a difference
 
oh, whoops, that's just me checking out the wrong branch
!!/master
 
12:28 AM
[ SmokeDetector ] SmokeDetector started at rev 95ee2fe (Thomas Ward: Changes to Post class, bodyfetcher.) (running on ArtOfCode/EC2)
 
!!/gitstatus
 
On branch deploy
Your branch is up-to-date with 'origin/deploy'.
nothing to commit, working directory clean
Restart: API quota is 18569.
 
I'll make a pull request if everything looks good
 
[ SmokeDetector ] SmokeDetector started at rev c44e0bc (ArtOfCode-: Fix index out of range bug) (running on ArtOfCode/EC2)
[ SmokeDetector ] SmokeDetector started at rev c44e0bc (ArtOfCode-: Fix index out of range bug) (running on ArtOfCode/EC2)
[ SmokeDetector ] SmokeDetector started at rev c44e0bc (ArtOfCode-: Fix index out of range bug) (running on ArtOfCode/EC2)
 
totally won't mess anything up if I just nuke the entire bodyfetcher queue right
 
12:34 AM
Not much spam ATM, so it shouldn't matter much
 
Restart: API quota is 18562.
 
!!/gitstatus
 
On branch time-in-queue
Untracked files:
(use "git add <file>..." to include in what will be committed)

bodyfetcherQueueTimings.p

nothing added to commit but untracked files present (use "git add" to track)
 
Ooh, we should strip those ANSI color codes.
 
eh
 
12:36 AM
[tag:dev-on-production]
 
flake8 doesn't like my code?
 
@angussidney best kind of development
@quartata flake8 doesn't like anybody's code.
 
So it likes the spaghetti code with no extra line breaks
but not my pristine code
so mean
 
Turn thee thrice ninety degrees anticlockwise, then flip thee head over heels backwards. Proceed to sacrifice thine firstborn child at the altar of Flake, and it shall forevermore accept thine code.
4
 
I don't even understand what it's complaining about
Ooh woops
OK that's valid
 
12:39 AM
@ArtOfCode don't forget to create three extra commits saying 'Thanks flake for making me put in those extra blank lines'
 
@angussidney That's exactly what happened
 
@angussidney "thanks flake" also works, as proven by my commits to the time-in-queue branch earlier
 
$ git log | grep [Ff]lake | wc -l
105
 
12:42 AM
Just over 2% of our commits are thanking flake
 
so, @Andy, what data was it you wanted? The file itself? I've got a file now, the data is still inaccurate because it hasn't averaged out over a long time yet, but it's something
 
We really really need to refactor findspam
 
@ArtOfCode I really just needed the file so I have something to play with. Accurate data, right now, isn't important
 
TBH, we really need to refactor Smokey for a lot of things. Python 3 support, re-write it to have better MS support (but keep it optional), use OOP, etc
 
@Andy gotcha. Same email?
 
12:50 AM
Yeah but findspam is the thing we modify the most
 
Yeah
 
And man test_post makes no sense.
 
@angussidney "thanking"
 
It generates why data for regex reasons without checking if it is a regex reason
And it tests on title username and body first and then checks if it needs to
 
On its way, @Andy
 
12:55 AM
@ArtOfCode Ooh time_in_queue would be really nice
combined with the rule performance data
 
@quartata that's... sort of the point of me writing that code
 
I know just giving encouragement
I hadn't noticed the branch yet actually
 
Looking at the queue code, it seems to focus on the sites which used to be spam magnets but aren't anymore
Maybe that should be the first thing we update once @ArtOfCode gets us some data
 
@angussidney let's let this run for a while to let the data average out, then we can look at it in a day or two
 
!!/location
 
1:02 AM
@angussidney ArtOfCode/EC2
 
@ArtOfCode, what do these numbers in the lists represent? Seconds? Milliseconds? I ask, because some of these are really small: u'health.stackexchange.com': [0.000291],
However, you can clearly see when each queue is full, because you get a tiny value like that as it calls the API
 
@Andy seconds. The minimums are tiny because we fire the API call right after adding a post to the queue if the threshold is met, which takes ~0.0003s.
 
That's going to mess with averages :)
 
@Andy should be okay, actually, because it's balanced out by the posts that were added in right at the start of the site queue
i.e. the max values
 
1:08 AM
@SmokeDetector k
 
:D I can set Windows' taskbar to behave like Ubuntu's
 
[ SmokeDetector | MS ] Pattern-matching website in body: Map of neighbourhoods for Canada by Tom on gis.SE
 
@ArtOfCode be on the left hand side?
 
@angussidney yeah, and hide itself automatically until I hit the left side with the mouse
 
CI on 92adc01 succeeded.
 
1:21 AM
so I now get the full screen dedicated to app windows
 
fp- by angussidney
CI on ec0b1fa succeeded.
 
@Art I've fixed the regex thing. Should I use destroy!, or check the return value of destroy?
 
@NobodyNada the former will throw an exception if it fails; the latter lets you display a more friendly error message
 
@ArtOfCode I think we'd rather throw an exception here; destroy shouldn't fail, so the detailed error logging seems more appropriate
 
1:24 AM
Restart: API quota is 18448.
 
Fair enough
 
@ArtOfCode Thanks for the merge
waits for spam
 
pushed, thanks @Art!
 
CI on 33ab6fe succeeded.
 
!!/pull
 
1:27 AM
[ SmokeDetector ] SmokeDetector started at rev 33ab6fe (ArtOfCode-: Make analysis script python 2) (running on ArtOfCode/EC2)
Restart: API quota is 18437.
CI on a46f747 failed.
 
how much you want to bet me it's flake?
 
That's cheating
It's always Flake
Also why do you guys have Travis and CircleCI
 
we race them
 
Spoiler alert Travis wins
 
CI on 13f9301 succeeded.
 
1:35 AM
Which one does Smokey use then?
And why doesn't Circle trigger on PRs
 
@quartata no, actually Circle wins more often than not
@quartata whichever wins the race
 
@ArtOfCode Really? Huh I guess this time was just a fluke
 
@quartata nah, it's being slow today
!!/pull
 
[ SmokeDetector ] SmokeDetector started at rev 13f9301 (ArtOfCode-: Ain't flake great) (running on ArtOfCode/EC2)
Restart: API quota is 18424.
 
math: min 0.000364, max 1284.30336, avg 463.017850933
mathoverflow: min 0.00045, max 1622.969337, avg 681.6722635
biology: min 0.000449, max 5.41631, avg 2.7083795
askubuntu: min 0.000298, max 615.706043, avg 99.8058600625
gaming: min 0.000429, max 1312.966034, avg 656.4832315
stackoverflow: min 0.000217, max 306.296978, avg 69.2360307909
mathematica: min 0.000299, max 1768.700984, avg 461.55807175
ru: min 0.000371, max 2614.520416, avg 1558.4881014
puzzling: min 0.000424, max 412.697364, avg 206.348894
 
1:40 AM
58 mins ago, by angussidney
$ git log | grep [Ff]lake | wc -l
105
106*
 
Bahaha. Line in the referring sites table for the Smokey repo view stats: 820915.rdfx.org 3 3
3 views, 3 unique visitors
 
that's your site for school, isn't it?
 
which means that some poor bastard reviewed my website, clicked a link, took one look and went "I understand nothing" and promptly left
 
lol
 
Man I wish there was a better way to test rules other than praying. I tested it on a post object I made myself but I have no idea if it will work on wild posts from the API
We should have a !!/test <se or metasmoke id>
 
1:48 AM
Also, I'm about to apply for work experience at <insert large-ish software company here>, nobody minds if I list my contributions here on my application form?
 
go for it, and good luck
 
Thanks
 
and if they come to us for references, I'll be sure to tell them how much of a pain you are ;)
 
I can guaratee that the spammers will say that I'm a horrible person :)
 
I'm probably applying for GSoC in a few days' time, and I fully plan to list my stuff here on that application too
 
1:51 AM
Good luck
 
ta
 
@ArtOfCode neat!
 
Applications open 20th, so I'll have to see if there's anything interesting there
(well, either interesting or not-boring enough that I'm willing to work it for a summer for money)
 
Random rant of the day: backlit keyboards can be bad, as you can see all the dead skin and hair underneath the keys in the lights
 
time for a can of compressed air
 
1:57 AM
Tried that, but it isn't very effective if you don't take off the keycaps
I've ordered a keycap remover online, it should arrive soonish
 
I wonder if it would be worth asking Pops about getting data on posts scanned and ignored by Smokey but then deleted later by red flags
 
@quartata I suspect that'll be a small number
 
I'd love to get false negative data but we can't get deleted posts
 
Might be worth having the data, though.
 
We could probably provide them with a list of every post/site ID we have on MS, then ask for all data since we started using MS
 
2:02 AM
@ArtOfCode not so much number as what and more importanly where. if site mods and 10ks don't know about smokey things might not get manually reported (and habitual spammers tere won't end up in our rules)
 
@Pops are you able to get hold of some data for us that we can't get? Ideally, we'd need site, post ID, and post creation date for every post that was deleted by red flags, for as far back as you can get it.
@quartata The majority of mods know about Smokey, if only because we had a pinned message in TL for most of a month
 
whoa whoa whoa not every red flag
 
@quartata Why not?
 
@ArtOfCode because then we'd have posts that we actually caught, but we didn't catch because we didnt have MS
 
@angussidney sure, but that's going a good two years back, and I doubt we'll get data for that long :)
 
2:06 AM
Also I was hoping for a more "foot in the door" approach in terms of CM diplomacy
 
"foot in the door"?
 
Asking for something small first.
Anywho how do we know if Smokey scanned it?
 
shrug I'd rather ask for what we could do the most with and then have Pops tell us that getting that isn't possible
 
Could have been during rollover or downtime
 
@quartata just assume Smokey scans everything. It's near enough true as to make no difference.
 
2:10 AM
All right.
When I get a chance I'm going to rewrite the magic eight ball with Keras
 
2:35 AM
i was asleep what's up?
@ArtOfcode ^
 
the sky pls don't kill me
 
punts @Riker into orbit
 
now the sky is down
 
@quartata I may want to reapproach your handling - i did a very thorough refactor to check that the Parent object is also a Post and such
must compare to my work
because i already had a branch in place for that refactor in the post object
 
If it works well in practise, why not leave it?
Don't just revert Quaratata's work because you've already done some
 
2:38 AM
No it's fine, I didn't know there was a parent prop in another branch
basically it's @Undo's fault
!!/blame
 
@quartata It's ArtOfCode's fault.
 
Meh that works too
 
haha
 
@angussidney It's a three line getter
 
@angussidney i did not mean revert
i meant improve
type checks, fix a major post init issue, etc.
not a HUGE issue
but first, pizza
though keep in mind I am also working on refactoring apigetpost to return a Post object too so...
... may result in some reverts if stuff conflicts, but ultimately not a major revert - just a different Post object approach slightly in unit and passing things
 
2:47 AM
!!/alive?
 
@quartata Of course
 
Quiet. Are things getting scanned?
 
@quartata huh, interesting, nice work. It's missing some things in some places - I've started getting anal about type hinting, etc. but other than that nice job
let's hope it merges cleanly ;)
(I have to merge into my branch to rebase)
 
Oh I would have done type hinting but didn't see it on any of the others. I value code consistency more than strict typing
 
@quartata uhm just one question
where're you setting parent= in the post object?
because I don't think that's implemented anywhere and would probably be implemented in apigetpost
which is what i was working on
 
2:53 AM
Undo did it in bodyfetcher
 
ah, nice, missed that, though that may have been me.
Glad it's there though :)
yawns I need more coffee
i thought that was in a branch, only/
@quartata I'm slowly becoming paranoid BTW about some PEP rules :)
 
Oh oops I didn't check blame on it
 
doesn't matter it's there
I hope answers are getting checked
@quartata it may be that it's not evil spam hour yet
!!/apiquota
 
@ThomasWard The current API quota remaining is 18244.
 
@ThomasWard Not a bad thing really. PEP is good
 
2:55 AM
okay then checks in 3 minutes
!!/apiquota
 
@ThomasWard The current API quota remaining is 18244.
 
MS says things are being scanned.
!!/rev
 
indeed.
@quartata Has it been checking answers in the past several hours? Since I pushed the fix to test answers proper
@ArtOfCode ^
!!/location
 
@ThomasWard ArtOfCode/EC2
 
2:58 AM
!!/test-a asdfasdf
 
> Would not be caught as an answer.
 
?
 
no i meant from the system ;P
test creates its own post struct and submits that for testing :)
 
!!/test-a aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa d
 
> Few unique characters in answer, repeating characters in answer
----------
Post - Contains 3 unique characters
Post - Repeated character: *aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa*
 
2:59 AM
i meant like the API and such, not the tests.
i.e. actual spam items
 
Just checking if FindSpam is OK. it was when I pushed
 

« first day (1217 days earlier)      last day (2603 days later) »