« first day (1205 days earlier)      last day (2631 days later) » 

8:16 PM
@ArtOfCode Yep, just a minute
 
@DavidPostill The OP may have modified within that 'grace period' before it puts a modification by OP as a 'new' record in the modification list
and it wouldn't show in the rev. history. I've run into it a couple times.
 
@Cerbrus thanks, but I know next to nothing about emojis...
 
@ArtOfCode Email in your inbox
I won't say anything else :P
 
...give it 10? :P
got it
 
great, lemme know when you've downloaded it
 
8:22 PM
also done
 
It's not like 22MB really costs anything to store, but ya know
 
@Glorfindel Allow me to provide some trivial: Sir Patrick Steward (aka. Captain Picard/Professor X/world renowned Shakespeare actor), will voice the poop emoji in the upcoming Emoji movie
 
I wonder if my auto backups are working or if the script is still broken
 
If it's the script I've been using, it's pretty much bulletproof if you set it up right
 
And, with that bit of trivia...we need Sir Patrick to lend his voice to our project.
 
8:24 PM
@Undo no, self-rolled. Needed to backup sqlite3 in the end, not mysql, so I sorta copied and patched from the one you linked and wrote some terrible bash and hoped it worked
I think I got the script itself working, actually, but cron refuses to run or something
 
Oh, well then no guarantees :P
 
...and now I can't get at s3 console
 
@ArtOfCode protip: sudo crontab -e works wonders
 
@Undo that's what I was using
 
secondary protip: unless you aren't working on root's crontab, sudo -u USERNAME crontab -e
@ArtOfCode so you broke it?
 
8:26 PM
@ThomasWard nothing said it was broken, it just... didn't put backups on s3
 
heh
 
J F
@ArtOfCode Like GitLab?
 
@JF "Production quality backups"
Use the right phrasing
 
yeah, last backup 21 feb, it's cron's fault
 
It's still from 2017 at least. That's not that old. :)
 
8:33 PM
old enough
 
I think I was around when 2017 started
 
there is a lot of spam in this database
 
decent amount, yeah
 
97 seconds to import the lot
 
62 seconds to import the lot
 
8:39 PM
What do you mean, I have to wait?
 
mountain of spam:
 
always wondered what SD did with it
 
I don't expect it to be implemented w/ the dev squeeze, but:
0
Q: Provide a way to reset - not invalidate - API tokens

UndoI'm running a service with ~300 write-authenticated access tokens. It's terrifying. As part of running such a service, I'd like to be able to cycle API tokens - so that if, heaven forbid, one makes an escape attempt and wasn't caught, it'd only be valid for a minimal period of time. /access-tok...

 
your tokens growing legs @undo
 
J F
@ArtOfCode But but the emoji are black and white ☹
 
8:46 PM
@Undo I think that zookeeper has done that before.
 
@JF they should be color where possible, black and white is the fallback
 
J F
@ArtOfCode How can I make it fallback? Setting font-family: 'Emoji One'; gets me this:
 
I just realized @Undo has the power to upvote his post about API tokens ... 300+ times in under a second
 
@Undo, I suspect the official response will be to not use no_expiry in the scope.
 
@Andy That... doesn't solve the issue
 
J F
8:48 PM
 
Now I just have to ask people to reauth every x days
 
J F
@ArtOfCode ^^
 
@JF that is the fallback
I have no idea how it works, but it's color where it can be and black and white when it can't
 
J F
@ArtOfCode So just only apply the font-family where emoji support isn’t detected?
 
@JF that could work, yeah
 
8:50 PM
@Undo Not disagreeing with you, but my guess is that's the answer. What you are asking for doesn't accomplish a whole lot from SE's point of view. If you've been compromised once, either you are still compromised, can be recompromised or are safe. I'm guessing the default assumption isn't "Oh they are safe now"
Resetting the tokens doesn't help in outcomes 1 and 2
 
It's not a bad idea from a security point of view, though - it's like changing your password, which you do (ha!) whether it's compromised or not.
 
Here's the real world example: When giving a dump to @ArtOfCode, it'd be great to be able to refresh tokens after zipping up the dump and before making it public, and be 100% insulated from any kind of leak that could come from that.
 
checks the dump for tokens
 
@ArtOfCode The solution in this case is to invalidate and force the user to retrust you, not to allow you - who were just compromised - to regenerate the password for me
 
This isn't a "I was just compromised" thing, this is a "Let me throw another layer of security there"
 
8:53 PM
@Andy you change your password whether you know it's been compromised or not, right?
 
@ArtOfCode Nope. I've used "password123" forever.
Hopefully that came across as asterisks for everyone :)
 
brb
 
i.e. even if you don't think it's been compromised, you change it to insulate against the remote possibility that it has been. This is the same thing.
@Andy yeah totally, mine is *********************
O.o had a moment there... Ran User.where.not(:api_token => nil) and it returned 246
 
@Undo Nothing wrong with that. Just saying as someone who maintains and API at work, I wouldn't let a developer reauth tokens for an end user.
SE may go a different way though. In which case, hooray for us!
 
@Andy Still don't see why you wouldn't.
Seems like it encourages the wrong behavior.
"Well, that looks a little suspicious, but... crap, it's a lot of work to invalidate all of these tokens so I guess I'll just ignore it."
 
J F
8:56 PM
@ArtOfCode [status-completed]
 
We force expire dates on our tokens too - but for up to a year. So, once a year a user has to say "yup, still trust this application"
@Undo Nah. I can invalidate tokens enmass
 
@JF nice
 
@Andy Oh, I can too. It's the aftermath that's hard.
 
@Andy so can we, but doing it will lose us a bunch of trust and a bunch of users
I could invalidate every token in the DB in about 5 seconds, if it came to it
 
@JF I know you were asking about adding an auto property to FlagLog. Planning to commit that?
 
8:58 PM
@ArtOfCode Reseting tokens doesn't solve that though. The token resetting is just something that has to occur. Whether through invalidate/explicit reauth or automated reset. The problem that has to be fixed is how the tokens were compromised
 
@Andy but they weren't compromised, that's the point
 
The issue there is you have to know they were compromised.
 
Or at least, they probably weren't compromised, with the same degree of certainty that I probably don't know your password.
 
@Undo If you don't, what's the point of reauthing?
 
22 secs ago, by ArtOfCode
Or at least, they probably weren't compromised, with the same degree of certainty that I probably don't know your password.
 
8:59 PM
@ArtOfCode Just told you above :)
 
@Andy Insurance. If I happen to dump the wrong database to Art, then we both know that any tokens there are already void.
 
J F
@Undo I will this evening.
 
Great, thanks
Idea is to protect against the dog-factor
 
Resetting tokens does solve that issue, though. I probably don't know your password, but you should go change it now just in case I do. The tokens Undo's holding probably aren't compromised, but it'd be nice to be able to change the lot at once just in case they are.
 
I'm not disagreeing with that.
But, by allowing a dev to reset a token, it allows them to hide potential compromises. "Oh, if we're compromised, I just reset everything." That's good...but it doesn't force you to think about the next step - securing it in the future.
 
Ah, so that's where we differ
 
For what it's worth, I did upvote it. It would be a nice feature to have.
 
This would give a developer an easy-ish way 'out' of a situation like that, yes.
 
I need to run though.
 
cya :)
 
9:03 PM
Yeah. It provides an extra layer of security for while you're not being compromised, but it potentially limits security improvements because you don't have to tell everyone why they have to re-auth and trust you again.
 
I suppose we're talking about the different levels of trust we give developers. I figure that if a developer isn't trustworthy enough to invalidate instead of reset, they may not be trustworthy enough to invalidate at all
 
@Undo was the migration to posts.deleted_at from deletion_logs.created_at retroactive?
 
Yes
 
excellent
 
Proud of that one, too :)
 
9:05 PM
certainly much easier to query for data analysis
Speaking of which, we've averaged 131 seconds to deletion with 3 autoflags
 
aye, I dug up the migration
 
sd why
 
@dorukayhan [:35864007] Body - Position 374-385: loss weight
 
9:26 PM
Stats: 131 seconds to deletion with 3 autoflags, 228 with one autoflag, and 5586 with none.
 
@ArtOfCode Thanks, that's useful :D
 
I was originally looking for a better method of detecting emoji support, and accidentally found a way to make support available everywhere...
 
That works
 
Would it be an idea to start utilizing folders a bit more on that repo?
 
J F
@Cerbrus We could also load Emoji One from a CDN.
 
9:29 PM
@JF charcoal-se.org is behind a CDN
 
[ SmokeDetector | MS ] No whitespace in body: PrivacyPrivacyPrivacy by Eray Tezc on stackoverflow.com (@dorukayhan)
 
J F
@SmokeDetector v
@ArtOfCode Like cdnjs.com i.e. an external CDN.
 
[ SmokeDetector | MS ] Blacklisted user: How can I automate the coloring of the dots(series) in a Scatter Plot? by Eray Tezc on stackoverflow.com (@dorukayhan)
fp- by Cerbrus
 
@JF shrug what's the difference?
 
Any objections if I move the emojifont files to /emoji/?
 
J F
9:31 PM
@ArtOfCode Fewer files in the repo? We use the ones provided by Emoji One instead of our copy.
 
@Cerbrus sure
 
J F
@Cerbrus We could also switch to CDNJS.
 
@JF Emoji One doesn't provide files
 
@JF that doesn't load the font
you need to have the @font-face declaration
and for that you need the font files, which I had to pull as a ttf out of a zip download, and convert into everything else
 
J F
9:34 PM
👍
 
Moved'm
 
make sure you change your @imports over too :P
 
If you move them together, the path stays the same, not?
 
J F
@Cerbrus The @imports in the AIM & FIRE CSS files
 
The paths in emojifont.css stay the same, yeah, but you'll need to change your own scripts' dependencies
 
9:36 PM
Were they added already?
 
JF had one in AIM at least
 
@JF: bad boy, you threw all lint rules out of the window there xD
 
J F
@Cerbrus Fixing.
 
Nah
I'm already on it :P
I'm making the "no emoji" class more generic
So fire / aim use the same css class
 
J F
Sounds good.
 
9:41 PM
[ SmokeDetector | MS ] Bad keyword in answer, bad keyword with email in answer, email in answer: How to build a U.S. credit history as a worker on a visa? by james on money.SE
 
In fact, we could make it a script that does the check adds the class, and imposts the css (<-- only if required)
 
tpu- by Jan Dvorak
 
^^^ Download link appears to be for cracked software ...
 
Hmm, I'm not so sure I like the "Emoji one" fallback...
The fire icon is hardly visible. It's black / white, and very thin lines...
 
10:17 PM
Restart: API quota is 8529.
 
J F
 
Nothing near that fancy
 
J F
@Cerbrus I was asking if it should look like that or if I should find a different image.
 
Oh, it probably should
 
@Cerbrus That poop is hardly visible.
 
10:27 PM
@DavidPostill My point exactly
Better than a empty square, but not by much.
 
@Cerbrus FIRE error:
Uncaught SyntaxError: Unexpected token o in JSON at position 1
    at JSON.parse (<anonymous>)
    at Array.socketOnMessage (eval at <anonymous> (:1:1), <anonymous>:789:25)
    at WebSocket.conn.onmessage (charcoal-se.org/userscripts/metapi.js?v=0.3.8:2:372)
trying to parse already-parsed JSON somewhere
 
Hm, remember you saying I need to double-parse something?
Maybe it wasn't quite that simple :P
 
Apparently not
It is returned as a string sometimes, but maybe not always
 
I'll just remove the parse for now, and log the thing in a final else
 
hang on, debugging
your problem is that double parsing JSON is your default case, but I don't think stringified JSON messages are ActionCable's default
maybe switch over to doing it just for those you know are stringified JSON
i.e. flag_log, deletion_log, feedback and not_flagged
 
J F
10:37 PM
@ArtOfCode Or don’t send stringified JSON from MS
 
AIM only ever does 1 parse
 
@JF we don't control that, it's down to ActionCable. We call broadcast with a hash, AC converts it to stringified JSON and sends that
 
It's meant for consumption by the AC JS client, which automatically handles turning it back into an object
 
[ SmokeDetector | MS ] Bad keyword with email in body: White Hat Hacker or Black Hat Hacker by Anton Olivier on stackoverflow.com (@dorukayhan)
 
and another one for you @Cerbrus:
Uncaught TypeError: Cannot read property 'site' of undefined
    at HTMLSpanElement.openReportPopup (eval at <anonymous> (:2:297), <anonymous>:369:28)
    at HTMLSpanElement.dispatch (ajax.googleapis.com/ajax/libs/jquery/1.7.1/jquery.min.js:3:4351)
    at HTMLSpanElement.i (ajax.googleapis.com/ajax/libs/jquery/1.7.1/jquery.min.js:3:328)
 
10:44 PM
@ArtOfCode I can't reproduce that one... There should be some logs nearby that might be of help?
 
Sometimes, d seems to be undefined [span.fire-button.fire-plain, context: span.fire-button.fire-plain] undefined
so yes, d is undefined
 
J F
@Cerbrus Fixed.
 
Fixed?
@ArtOfCode But why :P
 
J F
Sorry. I was unclear in what I was referring to.
 
Looks great @JF, only one small syntax thing. Need to figure out why CI didn't run on that.
 
10:49 PM
@Cerbrus what's this when you call openReportPopup?
 
The "FIRE" button
As in DOM element
 
@Cerbrus uh. Good question, then. .data("report") on that is definitely an object
so either it's not said object at the point the button is clicked, or this isn't what you think it is
 
@ArtOfCode Or the data call returns undefined
 
@Cerbrus but it doesn't, that's what I checked
 
I've gotta re-work the loading, any way.
That's odd...
 
10:53 PM
and... it's still the same object when the button is clicked, so this probably isn't what you think it is
 
Possible
 
J F
@Undo It did. Is CircleCI authorized in Integrations?
(it just couldn’t change the commit status)
 
Should be
It works on commits to master
 
Did font sizes on chat and SE sites just change for anyone else, or have I got a bug?
 
no-repro
 
[ SmokeDetector | MS ] Manually reported answer: Keyboard for programmers? by ArtOfCode on hardwarerecs.SE
 
fp- by Jan Dvorak
 
J F
@SmokeDetector f
@SmokeDetector why
 
@JF Post manually reported by user ArtOfCode in room Charcoal HQ.
 
J F
10:59 PM
@Undo this might help.
 
Well that got me a bunch of info. So, @Cerb, this is what you think it is - the FIRE button. $(this).data("report") is undefined when your call to openReportPopup fires. Don't know why. On the other (JSON) bug: at least one trigger is the feedback messages.
 
Maybe there's an error when the data is loaded...
 
@JF There we go, that must be it
 
@Cerbrus where do you set data("report")?
 
function loadDataForReport(openAfterLoad)
That is triggered on hover on the fire button. Theoretically, you could be too fast for the data to load
 
11:05 PM
Almost certainly
 
It's never been an issue for me, but maybe that's just my connection
 
I'd bet you that load takes probably half a second, and I'll guarantee you I'm faster than that
 
@ArtOfCode Me. (I don't even think it works on my phone...)
I can see some obviously
 
Fix: if the data isn't loaded when the openReportPopup callback fires, re-call it after a timeout of... 200ms?
 
Just not the really weird ones
 
11:07 PM
@ArtOfCode Nah, that's hacky
I've gotta re-work the loading any way
Cache -> reset cache on websocket message
stuff like that
 
!!/alive
 
Uh
Not again
 
Thanks for the report. Now go and revive Smokey.
 
oh, I killed it again
 
11:16 PM
No failover
What
 
Should we reboot it from SOCVR or is someone going to attach a debugger?
@quartata It's not dead, but it's websocket here is
 
Oh, right.
 
Okay, important matters first: FIRE now has a proper emoji fallback.
 
Why do we need emojis
 
11:19 PM
@JF: I'm now constructing a image from that github, if I don't have emoji support :P
 
It's not like Unicode doesn't have enough weird characters
 
@quartata Because these things matter
 
@quartata in general, to make shiny answers like this
 
The report didn't even make it into Metasmoke or SOCVR
 
Stars and boxes?
That doesn't need emojis
 
11:22 PM
nah, mostly phases of the moon
 
@quartata you can't see that? too bad..
 
also, ton of comets
 
Can't see it either
That's why we need a fallback
 
just upgrade your OS
 
It doesn't want to
And I don't want to re-install.
 
11:25 PM
My OS wants to be updated so hard it won't let me stop it from upgrading
 
Restart: API quota is 8179.
Restart: API quota is 8175.
 
!!/alive
 
@DavidPostill Of course
 
!!/location
 
@DavidPostill teward/ESXi
 
!!/alive
 
ded again
 
11:38 PM
@DavidPostill yeah, we got that from that report.
 
@JanDvorak That disappeared into the ether :/
 
the command still lingers
 
!!/alive
 
@NobodyNada Of course
 
@JF There is no NAQ feedback and it's gone anyway ... I suppose it is abusive ...
 
11:40 PM
Restart: API quota is 8146.
 
attempts to find half-decent ubuntu backup software
realises is a programmer and really just needs a tar of all my stuff
writes bash script instead
 
obligatory XKCD
 
... I culd really use some websocket messages now xD
 
@ArtOfCode Deja Dup is great.
 
11:44 PM
@quartata I've got a bash script now :P
@Cerbrus API channel?
 
I mean all you really need is rsync
 
tar -czvf /media/backups/current.tar.gz /home/me
 
I booted up a copy of SmokeDetector in my room for FireAlarm development to test !!/report. I accidentally reported a malformed URL, and the websocket died.
 
@ArtOfCode Yea
 
@Cerbrus ping
 
11:45 PM
[ SmokeDetector | MS ] Bad keyword in body, bad keyword with email in body: Hacking services by cyber on security.SE
tpu- by DavidPostill
 
J F
@SmokeDetector k
 
Excellent
 
Does !!/report take a "share" URL or the plain answer URL without the userID
 
either
 
11:48 PM
@SmokeDetector This keeps getting reposted.
 
@ArtOfCode So not that then.
 
Actually how was it even bumped?
 
@quartata Closed as dupe?
 
Oh, woops.
 
fp- by DavidPostill
[ SmokeDetector | MS ] Blacklisted username: Cef Browser Loads Recaptcha Very Slow by Dr. Madonna on stackoverflow.com (@dorukayhan)
fp- by DavidPostill
 
J F
11:56 PM
Would it be OK to rename this page to “Creating new spam checks”?
 
@SmokeDetector Can someone fp- that please?
 
@SmokeDetector f
 
Yay, functional caching!
 
@Cerbrus How many do you need? It has 4 ...
 
I just needed an update
 
11:58 PM
I'm thinking to remove the user from the blacklist. He has two reports both fp.
 
He's not blacklisted.
The username is blacklisted.
@SmokeDetector why
 
@quartata Username - Position 1-12: Dr. Madonna
 
@quartata That's what I meant.
 
Oh.
Who blacklisted it? Maybe there's a reason we're missing
 
A few months ago, someone with that username repeatedly posted a troll question -- like 10 of them
 

« first day (1205 days earlier)      last day (2631 days later) »