[ SmokeDetector | MS ] Bad keyword with email in answer, email in answer: phone broke. cann wallet is on phone . can i find my wallet elsewhere? by Henry Jokela on bitcoin.SE

tpu- by Mithrandir

[ SmokeDetector | MS ] Bad keyword with email in answer, email in answer, blacklisted user: How to retrieve unclaimed BTC for wallets whose private key is lost by Henry Jokela on bitcoin.SE

tpu- by thesecretmaster

Also, on another note, what does the FlagLogs table do, if not log autoflags?

@thesecretmaster that... is exactly what it does

Hm. Then the dump I have is lying.

Actually.... MS is lying, not just the dump

Autoflag stats on SO aren't 7353 (97.48%) TP and 153 (2.03%) FP are they?

that might be counting flags not posts

But that wouldn't impact the percentages, would it?

flags = posts*3

yes

maybe

shrug

So a better way to phrase that would be "Flags cast on tps <numbers>; Flags cast on fps <numbers>"

7346/150 flag logs, 3589/105 posts

[ SmokeDetector | MS ] Few unique characters in answer, repeating words in answer: When exactly does Steam mark someone as 'Online'? by yes on gaming.SE

tpu- by Glorfindel

[ SmokeDetector | MS ] Potentially bad keyword in body: Form Making Swift Xcode by westonadam54 on stackoverflow.com

fp- by Glorfindel

Probably a good heater on cold days...

Latest iteration of the site dash. I think I'm almost done :)

[ metasmoke ] ci/circleci failure on d4f21bf: Your tests failed on CircleCI

[ metasmoke ] ci/circleci failure on d4f21bf: Your tests failed on CircleCI

Looks awesome. Might be good to have a heading above the graphs saying "all reports"? Just to make it clear that those aren't all flagged reports

Also, how wide is your monitor in that screenshot?

I think that density is about the maximum. I'd look at the image size if I wasn't on mobile.

I managed to get a basic version of Smokey running, but without MS or GH. I am satisfied :P

@Undo ~1000px

Not pixels in the screenshot, css pixels

Someone give me the reason Alibaba is blacklisted. Follow up to that, should it be?

@Andy SO spam ring, ask ChrisF

@thesecretmaster I'd recommend doing a 2x2 grid of graphs rather than 3u wide

2 is about the most graphs you can fit in a row without them being ridiculously small

@Mithrandir got a link to that? I'm being asked by the other mods

I tried that at first, but the scaling got a bit weird. I can try again though.

@thesecretmaster you want two div.row each containing two div.col-md-6

Yeah, I got that, but chartkick doesn't exactly behave

e.g. stuff gets clipped

screenshot?

@ArtOfCode This was when I was first building trying to add those two graphs. I'll try again now.

and my other request: stick a line break after the form and before the stats paragraph

See also TL search

@ArtOfCode I remember why! It takes double the height if I do 2x2

uh

yeah..?

Then you don't see the counts at the bottom

Hmmm

I have a solution, sort of

maybe this:

Happy to push the tabs below the fold, that doesn't matter

and the two line charts there are the more important graphs

Well, I wanted the counts at the bottom because that's all "All Time" stuff, graphs are last n months.

I'm not sure where you're suggesting I put the pies?

2x2 grid

Oh

I like the graph-area size as it is though...

@thesecretmaster this would be good, if we can increase the height of the line charts some

try 200px

That's not 200, but I think it's a good size?

Looks good. You can get extra height on the reports line chart if you turn the legend off and shared tooltips on

Magic words for that?

somewhere in the library options to chartkick

hold on

Grr

Side note: I wish the segments of a pie chart could become links.

library: {"tooltip" => {'shared' => 'true'} } for shared tooltips

So the reasons chart could link to posts that matched that reason

add 'legend' => { 'enabled' => 'false' } to disable the legend

Or just legend: false

yeah, if that works

I guess and checked :)

I looked up the Highcharts API reference :P

Well, I think we've reached max informational density.

I have one more little change to make, then it's mergable.

gimme a ping when it's done

You'll want to run rubocop across it too

I... can't

Well, I can, sort of

Can we get this "failed" to say something like "were unable to be cast". Saying "failed" biases people negatively. Once it's said there, it requires explanation to convince people that it's not a general problem with the platform which needs to be resolved. As I said earlier, I don't think we should be saying that there at all, but at a minimum we should state it so it doesn't bite us in the rear.

I'll miss some stuff

Random thought: if there, for some reason (such as wonky instances), are two full Smokey instances running (with MS and everything), could that theoretically result in six autoflags being cast?

@Makyen I'll change it to "unable to cast"

@thesecretmaster Thank you.

@Mithrandir No, metasmoke will reject duplicate reports

@thesecretmaster how so?

Running rubocop -Da should do it

Yeah, but it corrects ~75 other random stuff that's in app.

@Makyen "We were able to cast automatic flags 7174 times (83.27%) and we were unable to cast automatic flags 1441 times (16.73%)."

@thesecretmaster huh, that's more than I'd expect

I assume that it's improved over time

I guess a lot of the people who signed up don't have accounts everywhere

@thesecretmaster that's... a good thing

But do you want that in my PR?

@thesecretmaster Sure. That's both more accurate and stays away from the "failed" word which can be negative.

@thesecretmaster sure. Tell you what, I'll run rubocop over the current code to clean that, you can merge that into your PR, and run rubocop on the results

k

@thesecretmaster done, no changes.

Um

Are you running rubocop from the metasmoke directory?

i.e. the same directory as the .rubocop.yml

@Mithrandir It includes a variety of things. Looking at the failure reasons (tooltip on "Fail") these are things like: no account; user had flagged another post within a few seconds of the autoflag and was a rate limited (many); user ran out of flags on the site for the day; Mysql2 errors; program errors (multiple types); user had already flagged that post. Basically, only things of interest to MS developers/the people flagging.

@ArtOfCode Yep

@thesecretmaster with/without bundle exec? Whichever it is, try the other one

if it still doesn't work, eh, I'll do it when it's merged

Why in the world does bundle exec make a difference there?? Grrr...

[ SmokeDetector | MS ] Bad keyword with email in body, pattern-matching email in body: How can i hack into his bitcoin account by stacey Gates on bitcoin.SE

tpu- by K.Dᴀᴠɪs

@thesecretmaster shrug

@SmokeDetector one more flag needed

@SmokeDetector free comment flags :P

!!/watch (?:1\W*)?512\W*605\W*1256

@K.Dᴀᴠɪs You don't have code privileges, but I've created PR#1672 for you.

@ArtOfCode has someone sent through a link to someone's profile for channels?

@angussidney yep

Yours I'm assuming?

indeed

@ArtOfCode BTW: The flagging page and the flag logs/failures tab has way fewer failures listed than those reported in just the "Stack Overflow: Dashboard". Are some failure reasons filtered out of the flag logs/failures tab?

@Makyen I don't think so?

huh, maybe it does

I'm sure there are more failures than 400

yeah, 14713

That's a big difference. Is the 14,713 just autoflag failures, or does it include manual flag failures?

@Makyen ah, there it is. 14713 is with manual. 331 just auto.

That's surprising

I think we actually filter some errors out before creating a failed flag log

which would explain the low count

@ArtOfCode Then the number used on the dashboard also includes manual failures, not just autoflag failures. Or it includes the ones you're filtering out from the flag lof/failures tab.

Yeah, we do

@Makyen cc @thesecretmaster

What is a manual failure?

Manual flags through MS can also fail

Ooohhhh

you want FlagLog.auto.failures for auto fails

OK

@ArtOfCode happens to me every couple days. Usually have to reauth MS with SE

not sure why

@thesecretmaster See the tooltip on those marked "Fail": metasmoke.erwaysoftware.com/flagging/logs?filter=manual

@ArtOfCode New dash pushed

@thesecretmaster Screenshot of what it looks like now?

@ArtOfCode One way or another, it would be a good idea for these to be consistent. Either filter those from the statistics, or show them in the flag log/failures tab.

[ metasmoke ] ci/circleci success on 983d342: Your tests passed on CircleCI!

[ metasmoke ] ci/circleci success on 983d342: Your tests passed on CircleCI!

I hope there's no dash JS that'll screw this up. I'm testing with no JS because webpack is broken on my machine.

None of our JS, I mean. I still have highchart and bootstrap and everything.

very nice

Nice. Ready to merge?

@thesecretmaster What's up with the selectbox at the top?

The styling looks off

@angussidney mac

in case anyone couldn't tell I'm half delusional due to what is probably a sinus infection

other half is inferno

Maybe disable the labels on the posts-by-reason piechart and use tooltips only?

@ArtOfCode If filtering, it would probably be reasonable to also filter "You cannot perform this action for another X seconds" and "You can only flag for moderator attention X times per day", as those are also common, known issues. In particular, the first is likely just a race condition with other user actions (unless MS is causing that on it's own).

and the third half is that is git issues

'and we were unable to cast automatic flags <x> times'

the only thing I have really to comment on is that it feels like there could be some more space between some of the graphs, it feels a little dense

@ArtOfCode Actually FF+linux. It looks fine in chrome.

@ArtOfCode READY!

@thesecretmaster oh, fair

Does that really matter that we tell them about failures?

It honestly has nothing to do with anyone but the developers/users themselves, not relevant to a site mod and may cause confusion

@angussidney might be used by more than a site mod, though

I'd use it

@angussidney 100% agree.

@thesecretmaster merged, good to deploy pending CI @Undo

A more appropriate statistic would be 'We cast x flags on y posts, out of z total posts caught (%)'

@ArtOfCode I'm going to say "Look at this thing I made!" in the TL after Undo deploys, if that's alright with y'all.

[ SmokeDetector | MS ] Link at end of answer: Skype two way screen sharing by Gutsyn Ralyzed on apple.SE

@thesecretmaster please do :)

@thesecretmaster ...I was literally about to ask if you wanted to do that or me :P

Be good to phrase it as an announcement of a thing that mods can use, in your own words, so we can pin it for a while

[ metasmoke ] ci/circleci success on 09d6676: Your tests passed on CircleCI!

A bunch of mods have asked for it, so I don't think that's unreasonable

@ArtOfCode if we had a per-user failure summary, I could see someone using it, but what's your usecase for a site-wide failure summary?

good to see if there's issues due to flag options changes

@ArtOfCode For what, outside of MS/CHQ development (I'm actually interested)? In other words, how are the things which we log as failures relevant to people who are not A) MS/CHQ developers or B) the user who was autoflagging/manual flagging and experienced that failure?

or an API problem

@angussidney me being able to look at numbers and say "something's wrong"

or lots of people lacking accounts there

or just that something is wrong

as long as it's phrased as "able/unable", I don't think it's a problem

Having it be "able"/"unable" is dramatically better. It still feels like information which could be confusing, but that may be a bias on my part left from "failures". I quite agree that it's very useful for MS/SD development/maintenance. I question if it's helpful/harmful to expose it to people not versed in the process/development.

I guess we can always wait and see how people react, then change it if needed