« first day (2423 days earlier)      last day (1390 days later) » 

4:04 PM
fp- feedback received
 
@Daniil I'm not sure yet. They only started showing up in the last 24 hours.
 
4:39 PM
[ SmokeDetector | MS ] Potentially bad keyword in answer (1): How to deal with nested useEffect? by Mefit Hernández on stackoverflow.com
iBug/Sandy: Executing automatic scheduled reboot.
[ SmokeDetector | MS ] Potentially bad asn for hostname in answer (1): Apex Specialist Superbadge Challenge 1 ✏️ by Nidhi gupta on salesforce.SE
[ SmokeDetector | MS ] Few unique characters in answer, potentially bad keyword in answer, repeating characters in answer (169): How to write to a file multiple times in java? by fghj on stackoverflow.com
tpu- feedback received
fp- feedback received on Apex Specialist Superbadge Challenge 1 [MS]
naa- feedback received on How to deal with nested useEffect? [MS]
[ SmokeDetector | MS ] Body starts with title and ends in url, link at beginning of body (140): May chieu nec np-ve305xg by Tân Hòa Phát on stackoverflow.com
 
5:08 PM
@ArtOfCode @Undo In the last several hours, we've had a number of SD reports which haven't resulted in a post record in MS. I currently suspect that for those posts the Markdown contained characters which the MS database can't hold, causing the creation of the record to fail.
I haven't caught one in the period of time when the report from SD would still be in the logs available to me. If the problem isn't obvious from the database scheme for the Markdown field, then if one of you has the time, it would be helpful to look through the logs to see if the reports are getting to MS and if there's any information as to why they are not being recorded.
Some of the SD reports are: 1, 2, one I back-filled.
Note that the one I linked above which I back-filled, was done prior to the bug being fixed which prevented the Markdown from being sent for !!/scan. So, the report from the !!/scan I did didn't contain the Markdown.
 
!!/watch tanhoaphatcorp\.vn
 
@Machavity Post 1: Already recently reported [ MS ]
 
@SmokeDetector Lies
 
[ SmokeDetector | MS ] Repeating characters in body (72): how use tqdm with loop lines file ✏️ by Abdulrahman Kamel on stackoverflow.com
 
@Undo Post 1: Already recently reported [ MS ]
 
5:13 PM
>.>
 
[ SmokeDetector | MS ] Blacklisted username, blacklisted website in body, potentially bad ip for hostname in body (190): How Do I Resolve Hp Laptop Speakers Crackling? by lily jones on superuser.com
tpu- feedback received
 
Yeah, no repro needed to diagnose that one
| title                  | varchar(255) | utf8mb4_unicode_ci | YES  |     | NULL    |                | select,insert,update,references |         |
| body                   | mediumtext   | utf8mb4_unicode_ci | YES  | MUL | NULL    |                | select,insert,update,references |         |
[...]
| markdown               | text         | latin1_swedish_ci  | YES  |     | NULL    |                | select,insert,update,references |         |
 
I assume that SO report was for testing?
 
yeah
trying to repro the markdown issue
 
Makyen/EC2-linux: Executing automatic scheduled reboot.
 
5:19 PM
Yeah, latin1_swedish_ci doesn't look right. :)
 
@Undo If you get a chance, I'd appreciate a deploy. I think I fixed GraphiQL but want to make sure the fix works on prod
Also if we know what's wrong with the magic deploy button, I'm happy to try and troubleshoot that as much as I can.
 
@thesecretmaster Did you try it? The last time I know of it being pressed, it worked.
 
Oh, no, I just thought Art said it wasn't working
 
It worked a day or two ago. At the time, I thought it was broken too.
Although, I assume that Undo is currently working on a change to the database scheme (i.e. the Markdown latin1_swedish_ci issue above).
 
5:25 PM
Oh huh it does look like it worked 2 days ago
I guess I can push the button then, worst case it fails.
 
[ metasmoke-deploy ] deploy started by thesecretmaster
[ SmokeDetector | MS ] Shortened url in body (44): Can we modify Thumbnails from the internet? by vchlaw on webmasters.SE
 
plz no kaboom
 
@Undo ??
 
5:28 PM
appears that the button doesn't actually trigger travis anymore :P
 
I enter Charcoal, and something's not working
 
[ metasmoke-deploy ] deploy started by thesecretmaster
 
<3 you magnificent bastards
3
 
pop (no kaboom, just a fizzle) :)
 
@Undo Why is markdown of type text and body is of type mediumtext?
 
5:29 PM
Was Undo testing or is there really something wrong with that question?
 
that is also an excellent question
testing
 
Go ahead, gonna go study ECG
 
I will be totally honest, this was the last way I expected it to be broken.
Do we just need to swap out some API keys or something?
 
Travis keeps screwing up the API keys
No idea why, they just stop working
 
[ SmokeDetector | MS ] Potentially bad keyword in username (1): How to capture NoClassDefFoundError by Clemens on stackoverflow.com
 
5:31 PM
Hmmmm
rude.
 
I'll lump it in with the deploy I'm about to do for the markdown col thing
 
thanks <3
Could we move the deploy script to a lambda or something?
 
Also the question title on SO is max 300 characters. Why do you have it limited to 255 in your schema?
 
Like the MS reboot button?
 
@thesecretmaster Isn't the problem
 
5:32 PM
@Dharman I thought the max title on SE was 150 Unicode characters.
 
Travis just forgets the token is valid; doesn't matter really where you call it from
 
Yeah, I mean skip travis, have MS kick a lambda, the lambda runs the deploy.
 
Lambda runtimes are limited to 30s
 
Oh really?
Huh, sad.
 
Oh
15min
Actually
 
5:34 PM
@Makyen Maybe you are right. I did not check. Is this documented somewhere?
 
That must have changed
 
I thought you could change the timeout
 
Maybe I was thinking of UTF-16 code units
 
I remember a little slider
 
Still though, I don't feel like rebuilding Travis' env setup stuff
 
Deploy running cc @thesecretmaster
 
yessir
 
However, that's not actually accurate. It is possible for a post to have a post length > 30k, if you're Shog9. [It was changed specifically to allow that for one of his posts.]
 
[ SmokeDetector | MS ] Link at end of answer (60): iRoot is not defined, 2nd time launch the app by Neranjan on stackoverflow.com
[ metasmoke ] ci/circleci success on a13e2b3: Your tests passed on CircleCI!
tpu- feedback received
 
5:38 PM
@Undo it looks sad
 
vv sad
 
Oh, the markdown PR didn't update the lockfile
 
I've got it
 
[ SmokeDetector | MS ] Link at beginning of answer (34): Babylon JS 3D File loading is not working by PixelPush on stackoverflow.com
 
5:40 PM
sd f
 
@Undo Isn't it basically just git clone https://github.com/Charcoal-SE/metasmoke && cd metasmoke && bundle && bundle exec cap travis_production deploy
 
[ metasmoke ] ci/circleci success on 040be90: Your tests passed on CircleCI!
 
@thesecretmaster Have to set up Ruby too
 
Lambdas have ruby IIRC
 
oh maybe
 
Post 1: Already recently reported [ [MS](https://m.erwaysoftware.com/posts/uid/stackoverflow/62753201) ]
Post 2: That does not look like a valid post URL.
 
Oop guess I don't know how that command works
 
[ SmokeDetector | MS ] Potentially bad keyword in body (1): Is a geared motor better for a DIY wind turbine? ✏️ by a deleted user on engineering.SE
 
5:44 PM
Is it stuck?
 
Seems kinda stuck
Hopefully it's not deleting everything
 
It moved!
 
Seems less broke now
 
Aaaaaay graphiql is back!
 
5:46 PM
@thesecretmaster The reason that you're adding at the end must be in double-quotes.
 
Ah, thank you
 
@thesecretmaster Thanks. That will help quite a bit. I'd been using a query I'd developed a year ago, but it needed to be updated a bit. Having GraphiQL will help quite a bit in doing that. cc @user12986714 you were wanting GraphiQL. It's now working, thanks to @thesecretmaster.
 
[ SmokeDetector | MS ] Link at beginning of answer (34): Arduino Analog pins interrupt by Sinusoidplus on stackoverflow.com
 
@Undo Thanks for taking care of the database issue.
 
Yep
 
5:53 PM
[ metasmoke ] ci/circleci: rubocop failure on d30aa90: Your tests failed on CircleCI
 
@Makyen Do you want the raw text for that one spam report that failed recently?
 
@Machavity I believe that's resolved (mostly). However, I'm not 100% sure which one you're asking about.
 
Wow there are so many tests that run when I push a commit now :P
 
[ metasmoke ] ci/circleci success on 86973da: Your tests passed on CircleCI!
 
Also, this is a known issue, but I feel like are tests are deeply nondeterministic and that is maybe indicative that they haven't been touched in ages and maybe that is indicative that we should revisit our testing strategy.
 
5:57 PM
@Makyen This
 
[ SmokeDetector | MS ] Pattern-matching website in body, potentially problematic ns configuration in body (97): Single and ready to mix, need some sexy or hot pics by Kelley Howard on stackoverflow.com
 
@thesecretmaster There's a random failure in the rails-tests. It appears to be an issue in the program which is setting up the environment for the tests. Basically, sometimes it just doesn't set up the database, or fails trying to do so. IIRC, the reported error is actually that it doesn't have auth to manipulate the database. However, when that is fixed, it then says that the database doesn't exist.
I've hypothesized that it may be an issue with tests being run concurrently. I haven't been successful in setting it to run only single threaded, so haven't actually tested that hypothesis. Not being successful in changing that is down to my general lack of knowledge of the language, tools and lack of a dev environment.
 
I think the problem is that minitest spins up multiple threads, then some exit early which kills their shared database connection.
From what I've looked at, it seems like minitest can run parallel, so it spins up a thread pool, but should only actually use one thread unless you tell it to be parallel.
 
@Machavity Ah, I'd missed that one. It's ironic that it's just prior to me reporting the issue. The underlying problem should be resolved at this point (database was configured such that the field holding the Markdown couldn't hold the characters used), so we shouldn't need the raw data.
@thesecretmaster That may be. However, what I was reading, months ago when I looked at this, was that 2 threads in parallel was the default, and you had to change the config to get it to only run one thread.
 
Oh that's strange. That wasn't my impression, but I could definitely be wrong.
Based on the docs here, it should only parallelize if you tell it to: guides.rubyonrails.org/testing.html#parallel-testing
And... even if I try to parallelize the way it says to in the docs it yells at me
 
6:15 PM
[ SmokeDetector | MS ] Post is mostly images (0): What are these underwater structures near 6°N 85°55'W by rezaul hamid on earthscience.SE
fp- feedback received
 
6:27 PM
@thesecretmaster I see "The processes are forked based on the number of workers provided. The default number is the actual core count on the machine you are on, but can be changed by the number passed to the parallelize method."
The tests on CircleCI are run on a 2CPU/4096MB machine, so I'd assumed 2 processes. I'd previously looked at changing that, but stopped looking when the CircleCI docs said "Note: A performance plan is required to access this feature."
However, looking elsewhere on the same doc page, it appears that may not actually be true, and that you can change it to small (1CPU/2GB) when using the free plan. So, we could try that.
@thesecretmaster Yeah, that's about how far I got on trying to change the parallelism in the test config. I also tried some searches and other suggestions, but I wasn't successful with any of them.
 
@Makyen Yes, but I think that's only true if parallelization is enabled. I was looking at the part where it says "To enable parallelization add the following to your test_helper.rb" which implies that parallelization is disabled by default.
The weird thing is that it's not only mysql errors
There are also just random errors
 
@thesecretmaster I'm not quickly seeing what started my impression that parallelism was default-enabled. However, that it was, was reinforced when I went looking for other people trying to disable it. I could be wrong.
 
Yeah, it feels like parallelism is enabled by default
On the other hand, I also feel like it'd be possible to disable it
 
@thesecretmaster I've only very rarely seen other random errors. By at least 10:1, if not more, the random errors I've seen have nearly all been the database issue.
 
I think it's super super machine / config dependant
I get about 50/50, but GH actions are almost exclusively other errors, while circle is almost entirely db errors.
 
6:34 PM
Well, we can try resource_class: small and see if that changes things.
 
[ SmokeDetector | MS ] Bad keyword in body, bad keyword with email in body (152): Java web server device upgrade by Yonathan Cohen on stackoverflow.com
tpu- feedback received
 
Actually
I can say with near certainty that it's not parallelism
Minitest has a kinda funny method that's called i_suck_and_my_tests_are_order_dependent!
And I added it
And it did not fix the problem
Oh wait, it gets weirder
I added a debugging puts before + after every test
And the db error occurs outside of a test
 
@thesecretmaster That's consistent with the issue being setting up the database.
 
The normal errors also occur outside of a test
Wait a minute
Could it be parallelism we introduced?
There are various places in MS that we spin up threads
It could be that minitest finishes before out threads do, then our threads get their DB connections pulled out from under them
This might be the fix
It's working locally for me now
I actually think I might've fixed it
I'm at 10 runs without an error now
 
6:50 PM
[ SmokeDetector | MS ] Potentially bad keyword in answer (1): How do i start the turn by turn NavigationLauncher from another activity by miked on stackoverflow.com
 
Gonna make this temp fix permanent and then see how that goes
 
@thesecretmaster I'm not sure which you mean by "this", but it's cool that it's working.
 
@thesecretmaster Wait, is there some part of Charcoal that is tested before rolling live?
 
@thesecretmaster Sounds good. I can get more info when you push it. My input isn't needed if you have something that's working.
 
@Makyen There are a couple places where we spin up threads in MS. I wrapped those in a function that says "wait for those threads to finish if RAILS_ENV == "test""
 
6:52 PM
@Makyen Besides, I'm really just thrashing around with not all that much familiarity with the language or tools. :;
 
The permanent version of that is using ActiveJob, which is what I'm about to try
 
Sounds good.
 
[ SmokeDetector | MS ] Potentially bad keyword in answer (1): How do i start the turn by turn NavigationLauncher from another activity by miked on stackoverflow.com
[ metasmoke ] ci/circleci: rubocop failure on 08b2096: Your tests failed on CircleCI
 
[ metasmoke ] ci/circleci success on 86973da: Your tests passed on CircleCI!
 
7:03 PM
I accidentally made the github action rubocop run with autocorrect on
 
[ metasmoke ] ci/circleci success on 2709ef0: Your tests passed on CircleCI!
[ metasmoke ] ci/circleci success on 24adb70: Your tests passed on CircleCI!
 
Gonna call that at least tentatively fixed
 
[ SmokeDetector | MS ] Potentially bad ip for hostname in body, potentially bad ns for domain in body (2): How to enable alt tag for the E commerce product page images in Magento? by Majid Khan on magento.SE
fp- feedback received
 
Although, while the Circle tests are happy, it appears that the github action test is unhappy
Which does not bode well
Although it's not a database issue this time
 
[ SmokeDetector | MS ] Potentially bad keyword in answer (1): How i scrape Flipkart Product comments review by php? by Raja Nepolion on stackoverflow.com
 
7:18 PM
@thesecretmaster It looks like there are communication problems between CircleCI and GitHub Actions. As far as I know, this is the first notification I've gotten from a GitHub Action. I'm tempted to just turn those notifications off, as I already get notifications from CircleCI about any test failures. I don't need double-notification. Although, the GitHub Action notice appears more about the communication failure than an actual test failure, so I guess it's not actually a dup-notification.
 
Yeah GH actions are a thing I added... today
Wanted to see if a different CI system would fix the CI errors or at least make it more reliable
 
Ahhh... I was wondering why I started getting extra email notifications.
 
Yeah, sorry, that's my fault
Although they're doing their job and pointed out a new error :/
 
:;
 
I'm very hopeful that it's a versioning thing
Although idk why it would be
 
7:27 PM
iBug/Coral: Executing automatic scheduled reboot.
[ SmokeDetector | MS ] Blacklisted username, blacklisted website in answer, pattern-matching website in answer, potentially bad ip for hostname in answer (262): If nothing sticks to Teflon, how does Teflon stick to frying pans? by techmartin on chemistry.SE
 
These notifications are not helped by GitHub's new UI for notifications. It used to be really easy to scan through notifications from different projects by just scrolling the page, dismissing all from a project with a single click(?) (IIRC). Now I have to click through to each project (wait for page load), select all, then mark as read/done. That's a serious PITA. It's a lot more clicks, dang it.
 
tpu- feedback received
fp- feedback received
[ metasmoke ] ci/circleci success on 252b717: Your tests passed on CircleCI!
 
@SmokeDetector nice that we're regularly passing on Circle
 
[ SmokeDetector | MS ] Pattern-matching website in body (96): problem graphing the loss of a neural network by Pabsby96 on stackoverflow.com
fp- feedback received
 
@SmokeDetector k Mod flagged since it's little more than a rant about FlipKart
 
7:39 PM
Alright, I think I've made a net improvement to MS
Now to do real life things
 
@thesecretmaster Yes, that's going to make a significant difference. There have been a few times when we've let things through that have been actually failing, but where the more likely to fail than not nature of the "random" errors has resulted in people missing the fact that there were actual failures.
 
[ SmokeDetector | MS ] Potentially bad ns for domain in answer (1): How do I put controls on screen? by Xafar on stackoverflow.com
 
My feeling is that the tests are largely useless anyways, but can't hurt
 
tpu- feedback received
[ SmokeDetector | MS ] Luncheon meat detected (40): Need help in android assignment (asap) by Prasad Vinay on stackoverflow.com
 
Generally, testing is great, but the MS tests haven't been touched in years AFAIK
 
7:44 PM
@thesecretmaster eh, they're useful for catching regressions
 
fp- feedback received
 
Yeah I really think I actually caught the test bug this time
 
[ SmokeDetector | MS ] Potentially problematic ns configuration in body, potentially problematic ns configuration in title (2): I have an issue in my website learnmitra.com, dropdown menu not working by DrGaurav Jangra on stackoverflow.com
 
Just got another DB error locally, realized it was in a call to Thread.new
just as I suspected :P
Idk where the best place to put this, but if circle fails again please ping me cc @Makyen @ArtOfCode
 
wilco
 
7:51 PM
I'm not claiming it's fixed yet, but when other errors come up I can fix them
 
Also, I just want to say, I think there is value in fixing the Github Action errors. They should be easier, and the problem should become apparent as soon as somebody is able to repro them.
 
[ metasmoke ] ci/circleci success on d411a40: Your tests passed on CircleCI!
 
@ArtOfCode Oh also I should mention that my changes should not break anything on prod. But, if they do break things, they'll break in a very bad way (i.e. no autoflags).
 
wouldn't be the first time
even in the last week...
 
8:03 PM
:)
 
[ metasmoke ] ci/circleci success on 4ea165d: Your tests passed on CircleCI!
 
My net contribution today: Added 3 passing actions to make the 1 failing action look smaller. Progress!
 
8:27 PM
Batman?
 
@thesecretmaster your changes good to deploy?
 
I hope so
 
teward/Osiris: Executing automatic scheduled reboot.
 
deploying
 
Exciting
 
8:28 PM
[ SmokeDetector ] SmokeDetector started at rev a7c724e (SmokeDetector: Auto watch of tanhoaphatcorp\.vn by Machavity) (running on teward/Osiris, Python 3.6.9)
Restart: API quota is 9642.
fp- feedback received
[ SmokeDetector | MS ] Bad keyword with email in body, potentially bad ip for hostname in body (58): Presidential Old Maid by presidentialoldmaid on boardgames.SE
 
and done
 
I hope autoflagging isn't broken
shrug
ooooo my rails console broke
exciting
 
[ SmokeDetector | MS ] Few unique characters in body, luncheon meat detected, no whitespace in body, no whitespace in title, potentially bad keyword in body, +2 more (438): mmnnnnnnnnnnnnnnnnnnnnnnnnnnnn ✏️ by vdvd12 on stackoverflow.com
[ metasmoke ] ci/circleci success on becfa6c: Your tests passed on CircleCI!
tp- feedback received
 
user435118
@thesecretmaster What did you change?
 
[ SmokeDetector | MS ] Link at beginning of body (42): Access issues within the same package? ✏️ by Fekete Attila on stackoverflow.com
 
8:42 PM
Many things: I added 3 after-push actions, one of which is failing, but to compensate I fixed one of the previously broken after-push actions. I also added GraphiQL back and fixed a bug in there. And finally I cleaned up some confusing code that lives near autoflagging stuff.
Also I wrote the code to resolve #713 but I forgot where I put it so shrug
Maybe I committed it by accident
that would be cool
Aw, I didn't :'(
 
!!/watch- xafargames\.com
 
[ SmokeDetector ] Blacklists reloaded at rev 2140b7e (SmokeDetector: Auto watch of xafargames\.com by Makyen) (running on teward/Osiris)
[ SmokeDetector | MS ] Potentially bad keyword in answer (1): PHP blocking website access by country ✏️ by wowzers on stackoverflow.com
 
user435118
@ArtOfCode Either you've missed my Keybase ping or you're busy or you're ignoring me (I hope it's #1 or #2)
 
[ metasmoke ] ci/circleci: rubocop failure on 465353a: Your tests failed on CircleCI
 
@SmokeDetector Wow, fast and informative :)
@thesecretmaster Aha! Found it.
 
8:51 PM
@Daniil it's not #1 or #2, it's better... it's #1 and #2
 
[ metasmoke ] ci/circleci: rails-tests failure on 465353a: Your tests failed on CircleCI
 
Noooooooooo
:'(
 
user435118
@ArtOfCode :)
 
@thesecretmaster Muahahahahahaaaa!
 
user435118
@SmokeDetector Shush Art
 
8:53 PM
Stappit!!
 
[ metasmoke ] ci/circleci success on b1ec8a9: Your tests passed on CircleCI!
 
Also @ArtOfCode Circle is angry at me right now, so could you send me the trace from that failed run?
 
Latest run is three green
the one before that was red, but that's not the latest
 
Yeah, the red one
Unless...
Yeah, the rails-test failure
 
rubocop or rails-test
sure
 
8:55 PM
thank
 
Haaaaa have fun
Error:

APIKeysControllerTest#test_should_revoke_write_tokens:

Errno::EADDRNOTAVAIL: Cannot assign requested address - connect(2) for [::1]:6379

    config/initializers/redis.rb:11:in `logging'

    app/models/concerns/websocket.rb:18:in `broadcast_event'

    app/models/concerns/websocket.rb:30:in `broadcast_destroy'

    app/controllers/api_keys_controller.rb:14:in `revoke_write_tokens'

    test/controllers/api_keys_controller_test.rb:28:in `block in <class:APIKeysControllerTest>'
@Daniil mostly harmless in this case, but general rule: don't reveal information you get from mod tools without the consent of everyone involved
 
@ArtOfCode Well, that's gross
 
yup
enjoy
 
[ SmokeDetector | MS ] Potentially bad keyword in answer (1): Who did something wrong in 2007-2008 financial crash? by Steve Estes on politics.SE
 
glad it's not me
 
8:57 PM
I was looking forward to a nice tasty DB error
@ArtOfCode It's ok, the level of redis use in MS is my fault, I should have to deal with the consequences
 
well I didn't say it :P
 
I love redis
It's so much fun
 
*gagging noises*
anything beyond SET/EXPIRE is me done
 
Got a lil side project I'm working on that actually manages to outperform it
(only under very heavy load, and with a tiny subset of the functionality)
(but I'm working on expanding functionality, and it should be possible to beat redis's performance with the exact same functionality)
My big challenge is implementing a doubly linked list
 
fp- feedback received
 
!!/watch- (?-i:M1cN569zzG2Y5M8q8)(?#forms.gle)
 
10:04 PM
@thesecretmaster Sounds like an interesting project. I hope it goes well.
 
10:21 PM
There's been no scan activity for 3 status pings. There may be a problem.
Recovered from BrokenPipeError: [Errno 32] Broken pipe
 
10:36 PM
@SmokeDetector Hmmmm... So, no activity in 3 pings isn't all that rare, even during the week.
@Makyen OTOH, it could have been associated with the BrokenPipeError.
 
iBug/Sandy: Executing automatic scheduled reboot.
[ SmokeDetector | MS ] Potentially bad keyword in body, potentially bad keyword in title (2): Are books in pdfdrive.com legal or illegal? by guest on law.SE
 
10:52 PM
fp- feedback received
[ SmokeDetector | MS ] Potentially bad ns for domain in answer (1): Thoughts on going open source and attracting contributors by Nuritzi Sanchez on stackoverflow.com
 

« first day (2423 days earlier)      last day (1390 days later) »