« first day (2601 days earlier)      last day (1513 days later) » 

tp- feedback received
@SmokeDetector Undisclosed affiliation. Author of linked youtube channel.
@ppwater User is blacklisted (11832105 on stackoverflow.com).
@ppwater User removed from blacklist (11832105 on stackoverflow.com).
[ SmokeDetector | MS ] Link at beginning of body (42): Is there other place download "Firebase Services"? by user924245 on stackoverflow.com
01:34
fp feedback received
02:24
[ SmokeDetector | MS ] Potentially bad ns for domain in body (1): Dualboot ubuntu and windows 10 help by c0nnc0nn on askubuntu.com
fp- feedback received
[ SmokeDetector | MS ] Potentially bad keyword in body (1): The value in the array? by Rivaldy Saputra Agus on stackoverflow.com
fp- feedback received
02:41
iBug/Lily: Executing automatic scheduled reboot.
03:10
[ SmokeDetector | MS ] Blacklisted website in body (99): Can someone help me understand what that means, please? Thank you by Bérénice on codereview.SE
[ SmokeDetector | MS ] Potentially bad ip for hostname in answer (1): why using support vector machine? by Nandkishor Blogger on stackoverflow.com
03:40
[ SmokeDetector | MS ] Link at beginning of body (42): What could be more optimized or which case is missing? by Durgesh Kolte on stackoverflow.com
!!/watch machinelearninginhindi\.in
@SmokeDetector tpu- Spam: This user's answers both heavily promote machinelearninginhindi.in, which, being written in Hindi, is clearly not relevant to the English-language SO audience. Further supporting this being spam and not an attempt at being useful, the last link in this post "for detail" is just to the homepage.
[ SmokeDetector | MS ] Potentially bad keyword in answer, blacklisted user (72): What is Reinforcement machine learning? by Nandkishor Blogger on stackoverflow.com
04:14
[ SmokeDetector | MS ] Link at beginning of answer (34): How to port COMPLETED libgdx game from Android to iOS by user3854967 on stackoverflow.com
[ SmokeDetector | MS ] Shortened url in answer (57): How can I train my dog to behave for grooming? by CHIRAG on pets.SE
API quota rolled over with 7210 requests remaining. Current quota: 19999.
stackoverflow: 1563
math: 1083
ru.stackoverflow: 648
physics: 587
superuser: 504
askubuntu: 453
electronics: 365
es.stackoverflow: 348
unix: 304
stats: 247
blender: 232
pt.stackoverflow: 224
tex: 222
serverfault: 203
english: 200
ell: 198
diy: 180
gis: 168
mathoverflow.net: 168
apple: 155
salesforce: 150
mathematica: 138
worldbuilding: 126
magento: 125
wordpress: 122
dba: 113
datascience: 104
scifi: 100
meta: 100
politics: 90
cs: 88
codereview: 85
hermeneutics: 76
puzzling: 75
drupal: 74
softwareengineering: 71
tpu- feedback received
04:29
[ SmokeDetector | MS ] Link at end of answer, potentially bad ip for hostname in answer, potentially bad ns for domain in answer, potentially bad keyword in answer (64): I need to redirect my website link from ravisah.in/index.php to ravisah.in by HTACCES file by ajay yadav on stackoverflow.com
tpu- feedback received
@SmokeDetector just the accepted answer with extra spam
[ SmokeDetector | MS ] Potentially bad ip for hostname in answer (1): Unexpected EFI system partition ID on MBR disk by Iuri Guilherme on superuser.com
fp- feedback received
sdc bisect kashidigital.in/
04:39
@ppwater Matched by kashidigital\.in on line 22778 of watched_keywords.txt
sdc bisect machinelearninginhindi.in
@ppwater Matched by machinelearninginhindi\.in on line 25980 of watched_keywords.txt
[ SmokeDetector | MS ] Potentially bad ns for domain in body (1): How to hide description on categories and tags? by Law Lessons on wordpress.SE
@SmokeDetector URL hops through clickbank.
tpu- feedback received
!!/watch- martweb\.in
[ SmokeDetector ] Blacklists reloaded at rev 263421d (SmokeDetector: Auto watch of martweb\.in by Makyen) (running on teward/Osiris)
[ SmokeDetector | MS ] Potentially bad keyword in answer (1): Skipping lines of code in Python? by dumbass on stackoverflow.com
tpu- feedback received
naa- feedback received
fp- feedback received
Makyen/EC2-num02: Executing automatic scheduled reboot.
[ SmokeDetector ] SmokeDetector started at rev 63b113a (Makyen: This is the MS instance, so needs to post feedback) (running on Makyen/EC2-num02, Python 3.6.12)
teward/Osiris: Executing automatic scheduled reboot.
[ SmokeDetector ] SmokeDetector started at rev 263421d (SmokeDetector: Auto watch of martweb\.in by Makyen) (running on teward/Osiris, Python 3.6.9)
Restart: API quota is 19639.
Makyen/EC2-linux: Executing automatic scheduled reboot.
[ SmokeDetector | MS ] Manually reported answer (94): Robin_stocks showing incorrect top movers by SEO Developer on stackoverflow.com
tpu- feedback received
sdc watch- nyc-movers-packers\.blogspot
@cigien That pattern looks like it's already caught by Bad keyword in username; append -force if you really want to do that.
05:51
[ SmokeDetector | MS ] Potentially bad keyword in body (1): Class Type Series, how to convert and train? by Weseng on stackoverflow.com
@cigien Post 1: Already recently reported [ MS ]
I'm confused. If nyc-movers-packers\.blogspot is caught, then why didn't this get caught?
@CodyGray 'https://nyc-movers-packers.blogspot.com/2020/12/new-york-city-movers-and-packers-nyc.html' is not caught by a blacklist or watchlist item.
05:54
Because...regex
sdc test nyc-movers-packers.blogspot
@RyanM 'nyc-movers-packers.blogspot' is not caught by a blacklist or watchlist item.
> Bad keyword in title, bad keyword in username
----------
Bad keyword in title - Position 0-9: nyc-mover
Bad keyword in username - Position 0-9: nyc-mover
Yeah just -force it
sdc watch-force- nyc-movers-packers\.blogspot
PR#5421 ("cigien: Watch nyc-movers-packers\.blogspot") opened by SmokeDetector
05:56
sdc approve 5421
Ok, but what am I missing with that?
Why not also include the ".com", @cigien?
@SmokeDetector Also, what? "Bad keyword in username"? That wasn't a username. That's probably why the post didn't get caught.
@CodyGray Not sure of the exact rationale, but that's what Makyen said. There are several of these, though there's no list being actively maintained that I'm aware of.
05:59
I see. Weird. Why would blogspot be different than anything else? If it's better to elide the TLD there, why not for other spam as well to maximize detection?
@CodyGray test checks it against username, title, and body
@CodyGray Blogspot owns quite a lot of the blogspot.whatever TLDs, but they're all in the same namespace, so nyc-movers-packers.blogspot.ru would work equally well as a link to that site.
@RyanM Ah, ok. Last time I had a similar issue, it was because it was caught by findspam.py as well. I forgot about that, thanks. I'll start greping in findspam.py for those.
Ah, so most of the time, what I really should be using is test-q or test-a.
You'll begin grappling with groping grepping? Sounds gripping.
fp- feedback received
Please Cody, get a grip on yourself.
@CodyGray yeah, !!/test-a is what I usually use if that's what I want to investigate
@tripleee Is there actually a difference between -a and -q? In other words, do we actually have different patterns for questions and answers?
fp- feedback received
@CodyGray very marginally; the main difference is that a question also has a title
@tripleee So, test-q is a combination of test-a and test-t?
@cigien User is not blacklisted (160091 on stackoverflow.com).
@RyanM ... why?
@CodyGray well not exactly; the chat commands are a bit clumsy because you can't pass in a post in exactly the format it is received over the API
06:06
@cigien They have been spamming "buttercms" links within the past year. But that last report we just marked as FP just unblacklisted them.
@CodyGray You'd have to ask checks blame ...tripleee, actually, was the last person to modify it.
Conveniently :-)
@CodyGray Yeah, I was just checking, thanks.
I used to recommend leaving off the TLD for .blogspot. URLs but I often no longer do; the evidence is strong that the spammers themselves don't exploit this avenue of obfuscation much
@cigien Thing is, I'm not actually sure that is the correct action. The user shouldn't be unblacklisted just because of a report on a really old post was determined to be a FP.
@CodyGray +1, it'd be nice if it didn't unblacklist for posts with a modification date before the blacklisting...
06:08
@CodyGray Not sure. They have an uncommon name, so maybe we could watch the name or something? I don't know.
@RyanM the code has gone through a bunch of refactorings so a superficial git blame is not always likely to tell you who actually wrote something
@cigien We can just manually add the user back to the blacklist, I believe.
@CodyGray Oh definitely, I don't know if we should though. But being unblacklisted for an old post being fp is strange.
@CodyGray yeah, that's the recommended course of action for now; I do agree that the usability could be improved quite a lot but the gains are really marginal
@tripleee I know, but in this case it was actually a meaningful change to its functionality. The rest was last touched by a refactor, though, so I don't know the original author.
06:10
I think I've mentioned this before, but... I'd love to have a private mod-only instance of SD that I could use to blacklist users, watch specific posts, etc.
There's so much that this tool could do for me that doesn't necessarily fit perfectly with Charcoal's scope.
@RyanM oh yeah well I can reveal that this was just "where do we already do this for PK" kind of reasoning, not a very principled change
@RyanM has found evidence of @tripleee's clear bias against the UAE. I'm sure that'll be used against them later, when they try to run for political office.
@tripleee ha, fair enough :-)
@CodyGray I can seek exile in Tokelau if it comes to that ... I think I have argued that maybe they should be removed from that list
06:13
@cigien User blacklisted (160091 on stackoverflow.com).
Wait, you don't need to have blacklista privs to do that? Or did @cigien get them while I wasn't looking?
@SmokeDetector User has been spamming over the last year, but was unblacklisted for FP on a post that's almost 12 years old, so adding them back to the blacklist.
Another reasonable option (other than blacklisting the user) would be watchlisting the "buttercms.com" domain. Only 1 FP out of 6 TPs: metasmoke.erwaysoftware.com/domains/53055
@CodyGray User blacklist can be done by any privileged user (including via tpu feedback)
@CodyGray this is an unprivileged chat command, the user blacklist is much older than the ones we require privileges for
06:15
@CodyGray It must be already ...
sdc bisect buttercms.com
sdc bisect buttercms.com
@RyanM Matched by buttercms\.com on line 17445 of watched_keywords.txt
@cigien Matched by buttercms\.com on line 17445 of watched_keywords.txt
Ah, yeah
So no real point in blacklisting the user then.
Other than it requires less privilege.
$ fgrep -n buttercms *.txt
watched_keywords.txt:17445:1582042549	Das_Geek	buttercms\.com
$ date -r 1582042549
Tue Feb 18 18:15:49 EET 2020
06:16
@CodyGray As long as you have a machine to run it on, setting up a personal instance of SD is fairly trivial.
@CodyGray Can you check and see if they've only been spamming buttercms.com? Then removing them from the blacklist makes more sense.
I have neither the machine nor the Internet pipe, unfortunately.
urgh, I think there's more buttercms spam, actually...
@CodyGray the EC2 where Halflife is running could certainly accommodate this if we wanted to go there
@cigien You could equally check via MS. But yes, that appears to be it.
[ SmokeDetector | MS ] Manually reported answer (batch report: post 1 out of 2) (94): Need to integrate CMS into existing React application by Pooja Rathan on stackoverflow.com
[ SmokeDetector | MS ] Manually reported answer (batch report: post 2 out of 2) (94): Is there a CMS with a layout/page builder that I can use with GatsbyJS? by Pooja Rathan on stackoverflow.com
@CodyGray Right, I realized as I pressed enter :p
tpu- feedback received
@cigien They've got 12 answers, all deleted. Most were spam for buttercms. Others were just boring old NAA ("I have the same problem").
06:18
@cigien User removed from blacklist (160091 on stackoverflow.com).
@RyanM To be fair... those questions... kinda called for it...
@SmokeDetector Might as well, since the only links user has been spamming are already caught anyway.
@CodyGray yeah, they weren't exactly good...
Not just that. It's a little unfair to accuse someone of spamming when they answer recommendation questions with... recommendations.
They are literally answering the question.
(The recommendations were relevant.)
[ SmokeDetector | MS ] Bad keyword in body, potentially bad keyword in body (96): Do you like atv rentals Oregon- United States by Ocean Breeze atv rentals on meta.SE
06:21
tpu- feedback received
!!/watch florenceoregonatvrentals\.com
@CodyGray That pattern looks like it's already caught by Potentially bad keyword in body and Potentially bad keyword in answer; append -force if you really want to do that.
Ah, a regular :-)
@CodyGray Wait, you're right, those answers seem relevant.
The issue, in my book, is the overly promotional nature: "ButterCMS works great with React! Add a blog/CMS to your React site within minutes" is very low on information, and high on marketing. If they'd written a more detailed answer and disclosed the affiliation, or if the site hadn't been repeatedly spammed elsewhere...I'd be more inclined to call it not spam.
06:23
Do you have some actual evidence of affiliation?
If so, then yeah.
Otherwise, the lack of details in the answer makes it a very bad answer, but I don't in general think it makes it spam.
For the purposes of moderation, heck yes, all of that needed to go. And is now gone.
In one way, I fear that if I were not here, those answers would have been nuked but the questions remain. That'd be even worse.
I don't, other than that it's been repeatedly spammed and the account posted those two highly promotional answers for the same thing and nothing else. Which is pretty strong in my book.
Your evidence of repeated spamming could also be construed as evidence of popularity and suitability of the CMS ;-)
"What's with all this spam for Visual Studio Code?!"
Well, one was already closed, and I'd have cv-plsed the other except I suspected that wouldn't be necessary :-)
I do try to differentiate actual fans of stuff from spam :-) I've removed at least one watch for a piece of popular software, and given FPs to stuff that has been spammed previously
I would have sent a cv-pls for the non-closed one if you hadn't deleted it
Probably a del-pls too, since the other answer was also terrible. I'd have hesitated to del-pls the other one, since it seemed to have some potentially useful non-spammy recommendations, but I'm not sad to see it go, either.
So we shouldn't tell Scratte? ;-)
@Mast the primary feature of the Python regex package we use which isn't available on MS is variable length look-behinds. For most cases where we use it, it's possible to rewrite the regular expression such that a variable length look-behind is not used and get the same, or very nearly the same, matched posts on MS. Sometimes this is relatively easy, but other times it's fairly hard.
For some regexes it's necessary to manually combine a couple of different searches in order to achieve something which approximates what the regex will do in SD.
06:35
How do you manage to get decent performance out of a system running tens of thousands of regexes in nearly real time on massive numbers of posts?
fp feedback received
!!/unwatch fiverr\.com
@Makyen No such item fiver\.com in watchlist.
[ SmokeDetector ] Blacklists reloaded at rev ac788e3 (SmokeDetector: Auto unwatch of fiverr\.com by Makyen) (running on teward/Osiris)
@Makyen fiverr.com is already watched with a somewhat more complex regex which does a bit better wrt. FP. This watch appears to have been added in response to a post where it was already caught by the existing watch.
8
06:40
@CodyGray the code merges all the regex fragments into just a few actual regexes, but still a good question
@Makyen Good note. That's happened twice now that someone tried to watchlist fiverr.com. Would be nice to have a better warning system in place, but probably too much complexity.
there were some measurements done in the past (by Art IIRC, or was it Undo? Or Quartata?) and there is some limit to how much the volume can grow before we need to figure out how to parallelize things, but I don't think we are particularly approaching that limit
I do not recognize the name Quartata
they were active here a couple of years ago but I believe they changed their user name to something else
a spaghetto I think (from the room owners, where this is the one I always wonder about before I recall who it is :-)
Ah, yeah, that's it. (Not that I recognize them under either name. I just confirmed that quartata is their earlier nom de plume.)
06:44
@tripleee "a spaghetto" is what they changed their name to.
anyway, the DNS queries are probably the main bottleneck, not the regexes
How useful are the DNS queries, really?
In my limited experience, I rarely see those as offering much if anything.
@tripleee Yeah, but those could be largely done in parallel, rather than serial.
[ SmokeDetector | MS ] Potentially bad keyword in answer (1): How do you make an Android loading icon in Illustrator? by temaprinter on graphicdesign.SE
there's a number of them, some of them are quite useful in some circumstances, like where a pharma spammer hosts a number of new domains on the same IP address
06:45
naa- feedback received
actually tpu
@SmokeDetector spam
[ SmokeDetector | MS ] Potentially bad keyword in answer (1): How to execute TestSteps using groovy? by Coders Camp on stackoverflow.com
tp feedback received
@CodyGray The more complex regex could be modified to throw up the warning, at the cost of not being searchable on metasmoke due to needing a variable-length lookahead
oh wait, it's variable-length lookbehinds that are the problem, isn't it...
@RyanM Why would it ever throw up a warning when attempting to add a new regex to the watchlist?
06:49
@SmokeDetector username matches on blog. It might be them. They have another non NAA answer with that link.
The "this is already caught" warning. It could be modified to trick it into triggering here by matching "fiverr.com" as the entire message, I think.
@SmokeDetector They also have a NAA with a YouTube link, and a legitimate answer with that link apparently provided for reference. I don't think this is intentional spam so much as a user who doesn't know how to post answers. But, for the purposes of SD, I am going to call this a TP.
Coders camp is spamming. There's a NAA on a youtube link where they're affiliated, and a FP on with the coderscamp website.
@RyanM so this one?
watched_keywords.txt:1603163661	Ryan M	fiverr\.com/\w++(?!(?:[^<]|<(?!\/?code>))*+<\/code>)
@CodyGray 3 posts with undisclosed affiliation? How is that not spam?
06:53
@cigien It's a confused user who doesn't understand our quality requirements.
[ SmokeDetector | MS ] No whitespace in title (70): MissingComponentException ✏️ by BitNovato on es.stackoverflow.com
As I said, for the purposes of SD, I'm going to call it spam. But I don't consider it spam for the purposes of addressing it.
fp feedback received
Sorry if that's confusing feedback to leave, given how it mixes concerns.
I am wearing too many hats!
@CodyGray Ok. I've left a comment on the answer that's not deleted.
06:55
@SmokeDetector still standing
@Yatin What's the advantage of comments like that?
sdc unwatch fiverr\.com/\w++(?!(?:[^<]|<(?!\/?code>))*+<\/code>)
sdc watch fiverr\.com(?=/\w++(?!(?:[^<]|<(?!\/?code>))*+<\/code>)|$)
@CodyGray If someone missed flagging it then they will
sdc bisect fiverr.com
@CodyGray I remove those comments from MS
@RyanM Matched by fiverr\.com(?=/\w++(?!(?:[^<]|<(?!\/?code>))*+<\/code>)|$) on line 25981 of watched_keywords.txt
And testing, this command should now fail:
!!/watch fiverr.com
@RyanM That pattern looks like it's already caught by Potentially bad keyword in body and Potentially bad keyword in answer; in addition, the regex contains an unescaped "."; in most cases, it should be "\."; append -force if you really want to do that.
excellent. problem solved.
typo notwithstanding.
06:57
👍
!!/watch fiverr\.com
@CodyGray That pattern looks like it's already caught by Potentially bad keyword in body and Potentially bad keyword in answer; append -force if you really want to do that.
Ah, nice.
How'd you make it do that, again?
Explanation: I allowed it to match "fiverr.com" followed by the end of the string (note the |$), which will match the tested string, but not any real post. It'd match a title or username ending in fiverr.com, but that's never actually happened.
06:59
tpu- feedback received
Ah, I see.
And yeah, even if that happened, it would probably not be a false positive.
@SmokeDetector that... doesn't make any sense
even from the point of view of trolling
But I was planning on changing my username to cigienfiverr.com
if it were a problem, I could add a (?<^fiverr\.com) in there, too, but right now I think that would just make it more confusing :-)
@cigien good thing it only matches on word boundaries ;-)
Ah, then I can make the change ;)
07:02
@CodyGray no, that's Milliner
Every single time, "Fiverr" reminds me of Watership Down.
@Yatin Trolling is supposed to make sense?
[ SmokeDetector | MS ] Potentially bad keyword in body (1): How do i fix Upwork clone sliding page problem (JS) by random_18 on stackoverflow.com
fp- feedback received
@SmokeDetector why
07:03
@Yatin Potentially bad keyword in body - Positions 6077-6087, 9836-9846: SEO expert
[ SmokeDetector | MS ] Blacklisted website in body, link at end of body, link following arrow in title, pattern-matching product name in body, potentially bad ip for hostname in body, +2 more (399): Buy @ >> nordicoilcbd8.wixsite.com/nordic-oil-cbd-de by Mertnenowoi on askubuntu.com
tpu- feedback received
@cigien I suppose it doesn't
Flagged
I've been having FIRE issues lately where I'm seeing "No metasmoke reports found for url: undefined" on some posts, like that last report. Refreshing the page appears to fix it.
Firefox 86.0a1, using Tampermonkey
I just experience intermittent severe lag with loading MS.
!!/watch nordic[\W_]*+oil(?:[\W_]*+cbd)?(?:[\W_]*+oi?l)?\d*+
07:05
I got the Enthusiast badge for visiting Ask Ubuntu for 30 consecutive days. Thanks keto spammers!
11
fp feedback received
Also, another bonus is that the regex now won't consume what's after fiverr.com, so it'll make a few of the other watches on specific fiverr usernames show up better
We don't have a Pattern-matching product name in title reason?
I'd have expected one, to match with *in body.
@RyanM I had that issue a few versions back. Not now.
Also, I'm on 84.0.1, how are you on 86a already?
@Mast I wondered if someone might do a double-take on that :-) I run nightly builds.
07:14
Of Firefox?
Wow.
Why?
@CodyGray Two reasons, one reasonable and the other very spacebar heating–esque: 1) I am impatient and like shiny new improvements, and they're generally almost as stable as the non-nightlies (props to Mozilla). 2) I always have way too many tabs open, which consumes a lot of system resources. The "update and restart" button provides a convenient, fast solution to restarting the browser, killing all non-pinned tabs, and reloading with the tabs lazily loaded when I click them.
How is #2 any different than just quitting and reopening the browser with the "reopen tabs from last session" (or whatever it is) option enabled?
That requires more clicks and waiting for the browser to quit
Also, when did "I am impatient" become defined as "reasonable"? :-)
07:19
It's all relative ;-)
@RyanM You're a programmer! Write a freaking script. You could do that in half the time it takes for a single compile.
But I don't need to, because I already have a working solution :-)
Mmm
Doesn't it take a very long time to compile Firefox?
I remember just a few years ago, it took > 12 hours to compile Chromium.
Also because I am a weird user and I like to inject my weird edge cases into the telemetry.
I have no idea, I'm certainly not compiling them myself.
I don't think the telemetry captures your unique weirdness
@RyanM Oh, what? Obviously I misunderstood your original comment.
07:22
[ SmokeDetector | MS ] Link at end of answer, potentially bad keyword in answer (62): Many youtube videos don't load but show a "loading" icon instead by temaprinter on superuser.com
Yeah... I had interpreted "run nightly builds" as you running a build each night.
tp feedback received
@CodyGray It probably captures that I constantly have 200+ tabs open, and the performance ...oddities that result.
@RyanM That's not even unusual. You've seen Catija's tabs, right?
@CodyGray Ah, no. I run the browser resulting from the builds that Mozilla runs each night (I think they're actually more than that: twice a day or so, anecdotally speaking)
@CodyGray I don't think I have
07:24
She's posted screenshots and counts in various places, including SOCVR, I think. She keeps an insane number of tabs going at pretty much all times. So did/does Shog9. So do I.
My difference is, I also split them up semantically into different windows. I'm not some kind of savage who keeps a hundred tabs in a single window.
but yeah whenever things get a bit laggy or I want my graphics memory back to run a game, I just update and restart, and I'm set :-)
For running a game, wouldn't it be even better to just quit, though, and only restart after finishing the game?
oh goodness, yes, mine are split into...currently 7 windows with 214 tabs.
RIP your workflow, I know.
@CodyGray ah but then I can't look up builds for my character :-)
07:25
Oh. I see.
Back in my day, we played games without cheating on the Internet.
Or even having an Internet.
@CodyGray you're the other kind of savage whose semantic structures are puny enough to fit into less than 100 tabs
I'm a reductionist! I just break them down into increasingly greater components.
In general, I built this system to be able to multi-task. I have 32GB of RAM and a 1080Ti. Currently vaguely contemplating the hypothetical not-yet-released 3080Ti as an upgrade, since all reasonable (read: not 3090) options have less VRAM than my current setup, and I don't want less since I already have issues.
in this browser instance, 23 windows, probably uh 500-1000 tabs ... but I have three other browsers
I'm limited on the number of browsers I can run. I also have 4 instances of Visual Studio...
07:29
actually many of those windows just have a few tabs so I guess 400-500
I should note that 214 is after a recent purge of nearly all the SE-related tabs...
@Glorfindel It is MS, or the flagging container, being rate limited by SE. The issue appears to be associated with MS accounts where the user has defined their flagging preferences and flagging conditions such that they are wanting MS to flag for them on every site, but where the SE user can actually only flag on a small number of sites.
I suspect that the accesses to the SE API are just done too rapidly, potentially without obeying a backoff at some point, as MS cycles rapidly through trying users which have no possibility of being able to raise a flag on the site where MS wants to raise a flag.
Going through the code where the accesses are done and looking for where there might be problems would probably be a good idea. However, we could/should also substantially alleviate the problem by having MS periodically determine for each user the sites on which they can flag, rather than just assuming every user can actually flag on every site where their flagging preferences and flagging conditions would permit.
In other words, if a user can flag is being determined at every attempt to flag using that user. While doing that is necessary, the pool from which the person is being selected contains a substantial number of users who can't flag on a large number of sites. It would be possible to eliminate these users from consideration for flagging on the sites where they don't have an account.
@Makyen Somewhat relatedly, it currently logs a failure for users who flag from FIRE on sites where they lack an account. It might be desirable for it not to count those failures, if they can be easily differentiated, as they likely add noise to the failure data (which I don't personally watch, but apparently others do).
@RyanM Yeah, the "failure" data has a lot of noise in it.
@Makyen Any idea why it didn't spot this title?
Matched by the following regexes:
wixsite\.com on line 195 of watched_keywords.txt
nordic[\W_]*+oil(?:[\W_]*+cbd)?(?:[\W_]*+oi?l)?\d*+ on line 25982 of watched_keywords.txt
@RyanM And then you wonder why stuff breaks?
@Mast Worried there aren't enough reasons there?
@SmokeDetector I award this one the "most creative attempt to shoehorn in a contrived reason that their site is relevant" award
@CodyGray On that post, no. But I like to understand what I'm looking at. Perhaps the filter isn't catching all it could be catching.
@RyanM Indeed.
07:38
@Mast Because the watch was added after the post was reported.
@Makyen So the title-version was added later than the body-version? Ok, that would explain it.
[ SmokeDetector | MS ] Blacklisted user (71): Increase icon size in hitomis/circlemenu by Appeteria on stackoverflow.com
fp- feedback received
@Mast No. The post was reported at 07:03. The watch was added at 07:05/07:06. This search will show you the timestamps and order.
07:43
@CodyGray User blacklisted (2700962 on stackoverflow.com).
@SmokeDetector Add user back to blacklist (was removed due to FP on re-report of early post)
@Mast uh, no; the watch for that domain was added after that post, because of that post
the title check did trigger on wixsite.com
I thought Pattern-matching product name in body wasn't governed by the watchlist?
@SmokeDetector ignore-
@Mast Oh, it was "pattern-matching product name in {}" which was being asked about wrt. the title? I haven't gone through the code for it, but the comment says # Pattern-matching product name: three keywords in a row at least once, or two in a row at least twice. I'd assume that it's requiring word-breaks around the keywords (a quick glance makes it look that way), so the title definitely doesn't match that criteria, but the body does.
07:55
[ SmokeDetector | MS ] Bad keyword in title, potentially bad keyword in body, potentially bad keyword in title (100): Where can i avail Mulesoft Training in India? What are the career opportunities with Mulesoft Certification? by SK Trainings on superuser.com
tpu- feedback received
sdc bisect Mulesoft
@Yatin 'Mulesoft' is not caught by a blacklist or watchlist item.
sdc bisect SK Trainings
@Yatin 'SK Trainings' is not caught by a blacklist or watchlist item.
07:56
@Makyen I must say that I have had backoff issues as well lately where I'm pretty sure the backoff is handled (and I have a key and such). Sounds a bit like: meta.stackexchange.com/questions/229958/…
sdc watch SK Trainings
@Yatin You don't have blacklist manager privileges, but I've created PR#5422 for you.
PR#5422 ("Yatin: Watch SK Trainings") opened by SmokeDetector
@Makyen Ah, thank you.
Too many FPs for "Mulesoft".
"Mulesoft Training" seems to already be caught
!!/bisect Mulesoft Training
07:58
@CodyGray yea it is a software or something
@CodyGray Matched by mulesoft\s+training on line 16327 of watched_keywords.txt
@SmokeDetector :O
@Yatin Is there a reason you're not allowing different whitespace versions like our standard?
@Yatin this watch will have FPs... because of sklearn library in python

« first day (2601 days earlier)      last day (1513 days later) »