@djsmiley2k Floor was put down in sections, but cabinets seem to be on top of them

I can lift a cm or two the edge that has nothing near it (under sink), but the bit that has the side of a cabinet what looks like on top of it I cannot budge

So yeah

Currently rotating newspapers while running a dehumidifier (which hasn't produced much water) and have just added an electric fan heater to the mix

Builder is going to drop by tomorrow morn

Or at some point during the day

@allquixotic I'm seeing under 600 MB total res

split between two processes

@djsmiley2k how old are your spawnlings?

@Bob oh okay. the virtual was alarming :P

you wanna update FF or should I?

actually a general apt update && apt full-upgrade is probably in order

I can zfs snapshot it first if you want

@allquixotic maybe in a bit, unless it's urgent

@allquixotic firefox is a custom package at the moment

if you update it you'll have to update ghostdriver too

ah, okay, I'll do it then

ok, I'll play around with it

I snapshotted since what I'm about to do could break stuff

usn.ubuntu.com/usn

welp, the Ubuntu patches are out for the kernel

I'm going to reboot into the HWE kernel, which is 4.13, supported until August 2018; that gives me a few months after the release of 18.04 to upgrade to the new LTS

@allquixotic export too if you can :P

@allquixotic you'll just want to download the linux x64 tarball, extract it over existing, and do the same with ghostdriver

that should work

don't delete the profile

also wtf, every other message in the RU chat is getting flagged

wait what

Oh, Russian chat

> Duplicate. Due to the huge number of alarms in the chat, the results of the vote are canceled (the reason is whether it is connected with the chosen owners or not, but the chat began not to meet the requirements for content placed on the Stack Exchange platform). The future of this initiative is subject to further discussion by the community.

apparently 4 people flagged that message as inappropriate

(the original Russian)

o.O

the guy whose post was flagged is an SE employee: ru.stackoverflow.com/users/6/nicolas-chabanovsky

it sounds like they were trying to work with the people in the room to get around some of their frequently disruptive remarks, and when that failed, they sound like they're going to either kill their chat permanently or (maybe?) the site?

idk if it's about the site or the chat because the translation is not great

I think what happened to SciFi chat might happen to them too.

what happened to SciFi chat?

their main chat room got killed

I only learned about the drama a year later.

scifi.meta.stackexchange.com/questions/7517/… Read more.

@jokerdino is the "37,000 users" comment anywhere near reality? that seems impossible to even do

I've seen Twitch chats with 10,000 people; the comments scroll by so fast that you can't read them

I think the room was timed out so there maybe 37000 viewers.

@allquixotic yeah, I don't open Twitch chats.

urgh the old 'that offends me please stop' thing

but i refuse to leave

so the question comes down to 'is that ok in the public domain?' which I have no idea, as I haven't seen the messages, alas

I also don't care that much :D

:D

due to my special powers, I dug the things and yeah.

I think it was more of a recurring problem.

I think when it gets to the point that SE employees are spending significant effort banning people or policing the content of chat, they just shut the room down instead of dealing with it

and there in lies the problem of why this will never get dropped, never get forgotten

and I don't blame them

because any new user is going to ask 'So.... what exactly WAS said?'

they have work to do

yup,

agreed

@djsmiley2k well, people are curious. we like to look on at trainwrecks :P schadenfreude

@djsmiley2k any new user wouldn't know about the old room because they have a new room with a clean slate.

@jokerdino I'd never heard of Mos E until someone brought it up in arcaqe...

yup

May of fixed the dishwasher, I just don't know anymore

it's currently draining

@djsmiley2k Woo!

What did you do?

smack it in the head

took the pump out

found rice.

looked angryily at wife

cleaned it out

@jokerdino Percussive maintenance, you say :P

put it back together

@bertieb the best kind

Ahh, so rice was the culprit

so, input was malformed

@jokerdino Did it happen again?

@bertieb no, that was the final episode.

"year later" wasn't accurate I see.

also, I guess I mistakenly linked to the previous episode.

@jokerdino Ah no problem at all, was just something which wasn't on my radar at all so trying to get it all straight in my head :)

Welp, I'm caught up. tl;dr, people repeatedly failed to Be Nice I guess?

@bertieb except in Russian this time

@allquixotic Ahh, oh dear. I guess periodic reminders are needed

@djsmiley2k Googled for other similar incidents as per your good suggestion; tl;dr floor is (probably) knackered with a capital 'f'!

yeah :/

Floor

I found the same thing @bertieb :

Ahh :P

but if you're lucky and you got most of it up, it may just warp slightly at the edges and not be too bad

^ Aye hopefully! Will see how it goes (less water than the poster there but still enough to have actual moving water under boards, not just dampness). If not will be checking home insurance and asking builder hopefully tomorrow how much parts+labour would be for new floor

The exposed bits are drying up not too bad, but it's the bit further along that's going to be troublesome perhaps

Trying the newspapers to wick/draw it back but we'll see

@jokerdino twitch chats literally crash my Z2760 tablet

NotLikeThis

granted, it lags horribly in the browser so I had to grab a store app to watch videos on it

it just happens to crash that app on particularly floody channels :P

lol

wow this website is super slow considering it is on cloudfront

Edit: no oneboxing I guess. I can still squeeze water from further along the floor. Erk.

ugh

how did water seep there?

@bertieb you get oneboxing if you directly link to the image, not the page containing the image

@allquixotic Ah, I was 50/50 on which one to try :P Cheers!

but it's an MP4

you can't onebox an MP4 in chat

(as everyone with limited data connections much appreciates)

@jokerdino Water leaked from under sink ± dishwasher drain overnight

@allquixotic me_2years before

@allquixotic Aye fair enough, I wondered about the size!

there is a gifv version i.imgur.com/JkS89Ov.gifv

(source is 2 meg, I could probably make it smaller with some ffmpeg jiggering)

Or not, as it's about 4x the size...

ffs

@bertieb plumbing issue?

been on hold for ages

its the same 30 seconds of music looping forever

after 2 days, it will be "closed"

@jokerdino Probably, I have a now-visible drip but nothing fast enough to cause that volume of water

Mind you, overnight is a long time...

yeah, agreed

Well, I have a cropped gif but it' still pretty large so I'm gonna hang fire on including it for the time being

Seeing as those who want[ed] to can [/have] seen the imgur link

Well, unless 1.6 meg isn't terribly objectionable... (JG isn't tethering anymore!)

Hah! The electric fan heater managed to trip the thermal cut out in the extension cable (too much current going round the metres and metres of cable I guess)

The sockets on the extension have a wee 'reset thermal cutout' button

Have unwound the cable so it is less inclined to heat

@bertieb not terribly, no, but when Bob was on mobile in the past he'd occasionally complain about images of that size

@allquixotic meanwhile people on limited data connections cry about huge GIF oneboxing where MP4 would be 1/10th the size (or even smaller)

@allquixotic I was complaining about the 10MB+

which included some 70-100MB ones

@Bob for a given PSNR, and assuming a good codec, yeah... but MP4 is just a container, and could in principle contain a terrifying amount of data, which if your phone starts downloading while you look away, could be tens or hundreds of megs before you notice

if chat opened the floodgates to oneboxing video, that stuff is a lot easier to abuse than GIF, imo... GIFs generally have size maximums that are fairly reasonable on image posting websites

@allquixotic ...imgur does 100MB+ GIFs

limit is 200MB actually

@Bob I seem to recall them not at some point, but they could have changed it... definitely recall pretty restrictive limits on other sites though, not sure of the specifics

and if I wanted to post malicious bandwidth-sucking oneboxes I could easily host a 1TB GIF

I occasionally get errors trying to post to puush with big gifs

you're deoptimising the common case out of concern for abuse that's already possible

if anything, MP4 is safer: browsers won't (?) auto-play/download them

meanwhile GIFs always download/play the full file without warning

wait, didn't they make it so all oneboxed content is no longer direct links to the content, but are now hosted on i.stack.imgur?

@allquixotic nope.

how are they getting around the HTTPS/HTTP browser security thing then, if you link an http image?

huh, I now have <img class="user-image" src="http://website-thumbnails.informer.com/thumbnails/280x202/n/neverssl.com/pn‌​g" alt="user image"> in my markup

ah, right, mixed passive content is allowed

Also, that gif above ^ could have easily been 1 TB

and your browser would've happily downloaded it

at least until it OOM'd

literally right then the insecure content warning showed up on my "lock" icon (when I posted the NeverSSL pic)

so yea

the attack you describe would be pretty easy right now

I guess they're not oneboxing video because they don't have anyone doing anything to chat code, not so much that it would be a bad idea

(though, I actually can combat it: get Cavil to HEAD the file, check Content-Length, and delete/flag the message if too big)

If it's a stream without a Content-Length, well... Cavil's on a fat pipe, he can download, say, 20 MB and nuke the message before most mobile users get there.

@Bob and server-side they could download the file and re-host it if and only if it's smaller than some arbitrary size... they couldn't check the length and allow it through if it's small enough because the size could change after the fact

@allquixotic that's a lot more effort, and possibly a licensing issue

@Bob if it's a license issue then they have the same issue whenever anyone uploads anything to the mainsite or chat right now; I can go find a copyrighted image I don't have the right to reuse and have it hosted on i.stack right now, and probably have done that many times already

@allquixotic I think it's the difference between the user explicitly performing the upload vs an implicit copy

@allquixotic I can update Cavil's browser now if you want?

I don't think the code would be especially terrible, but the storage concerns could possibly (?) be an issue - not sure how much storage imgur gives stack

@Bob sure; though his uptime is going to be short because I'm getting ready to reboot the physical box for Meltdown and Spectre (new microcode and kernel from Ubuntu and it isn't live patchable)

btw, if you want to laugh, they had functionality regressions and crashes in their initial patchset, which was released on January 10, not January 9

so technically the correct Meltdown fix for Ubuntu didn't arrive until 1/11

@allquixotic yea, if you're gonna nuke uptime anyway :P

usn.ubuntu.com/usn/usn-3522-3

@allquixotic you rebooting right now?

it's been a good run, 4.4, but all good things must come to an end

or after I update? :P

@Bob after you update

do an apt update && apt full-upgrade while you're in there?

k

I'll do the same on my other containers so I'm fully patched

@allquixotic btw if memory usage ever becomes an issue you could just quota the cavil box

and I can tune firefox to use less (e.g. disable e10s)

right now it's just a default config

I just remember glancing and noticing the VIRT in top at some point and going "whoa"

didn't actually investigate

but it was around 20 gigs

@Bob that a bunch of movies?

              total        used        free      shared  buff/cache   available
Mem:           125G        1.6G        124G        3.0G         13M        124G
Swap:            0B          0B          0B

@allquixotic the java webdriver is actually using more than firefox proper

gah, I really need to get cavil to bump the bot room weekly

@jokerdino, could you unfreeze chat.stackexchange.com/rooms/40974/bot-overflow?

> For clarity, the Canonical Livepatch Service is only available and supported against the generic and lowlatency GA kernel flavours for 64-bit Intel/AMD (aka, x86_64, amd64) builds of the Ubuntu 16.04 LTS (Xenial) release. HWE kernels are not supported at this time.

looks like I'm not going to HWE after all

@Bob tada

glad I researched that before rebooting into 4.13

@jokerdino thanks

@allquixotic Firefox is at 2 GB virt... Java is at 35 GB virt.

typical java :\

@Bob that's a per-container statistic, but yeah, having the cavil box only use 1.6 gigs is nice

@Bob you trimmed your active chat room list?

@jokerdino idk, I think chat reset a few times

@Bob IIRC bloated virt just creates a lot of noise in the page tables, but is generally not a big deal unless it's actually mapped

oh wow, I can see Bob's profile and gravatar in chat now!

@allquixotic yea, that's actually exactly the issue I had with subgit running on java on windows

he isn't in as many chatrooms as he has browser tabs!

large virt => over 1 GB of page table entries locked in RAM

it was ridiculous

@Bob ouch

compare with subgit closed, ~22 MB (kB? whatever unit rammap used)

@Bob thought so, didn't think you are the kind of person to voluntarily leave chat rooms.

@jokerdino or voluntarily close browser tabs

yeah, that's even worse

at some point, I heard he had 4 digit tabs.

@allquixotic looks like we're running firefox 55 right now

I'll chuck the current 57 on there

that helps

@jokerdino yeah; he was approaching five digits, but when he hit 9001, his browser considered that there was "no way that [number of tabs] could be right" and tripped an assertion, crashing his browser and deleting all his tabs

:D

How long would a browser take to cold start with so many tabs?!

interesting that free and similar stats are container-private... kind of a good thing for untrusted containers, but there's plenty of reason to be wary of untrusted containers before Meltdown; now I'm worried about the viability of platforms like Joyent Triton

@jokerdino if it tries to load the content in each tab, probably hours

IIRC Firefox only loads content when you click on tabs

sean@cavil:~/dev/firefox$ ls firefox
application.ini     gtk2                libnss3.so       minidump-analyzer
browser             icons               libnssckbi.so    omni.ja
chrome.manifest     icudt58l.dat        libnssdbm3.chk   pingsender
crashreporter       libfreeblpriv3.chk  libnssdbm3.so    platform.ini
crashreporter.ini   libfreeblpriv3.so   libnssutil3.so   plugin-container
defaults            liblgpllibs.so      libplc4.so       precomplete
dependentlibs.list  libmozavcodec.so    libplds4.so      removed-files

I think I can nuke that

ffprofile is a different directory

@Bob nuke with abandon; I have a zfs snapshot of your entire container from this morning

before we touched it

Bob certainly had a presence in so many chat rooms, so he ought to have loaded those tabs at least.

@jokerdino about 5 seconds. firefox heavily optimised the "thousands of tabs on load" case in ... 54?

O.o why does mine take 30 seconds with just 5 tabs?

get a SSD

@jokerdino HDD?

jinx

oh I see

sean@cavil:~/dev/firefox$ rm -r firefox
sean@cavil:~/dev/firefox$ tar xf firefox-57.0.4.tar.bz2

I would like some more cheat codes.

whoops

s/ghostdriver/geckodriver/

when you said ghostdriver earlier I was about to ask a question about why you were using that, but then I decided to keep my mouth shut because there was probably a good reason

but yeah, ghostdriver typically only drives phantomjs

or slimerjs

ok, I have geckodriver 0.19.1 now, from 0.19.0

...what else

has there been a replacement for tab groups yet?

there is some crap called containers but I don't like it

Selenium/Java might need an update if you're using Java as the driver

if your project is mavenized, edit pom.xml

@jokerdino I use containers :( they're nice

i couldn't get them to work

@allquixotic probably not - the new geckodriver has minimal changes over the previous one

I have a work website that freaks out if you have certain cookies generated by other applications on the same domain (for that matter, it freaks out if you navigate away from a page before all resources on that page have loaded) and I stick that in a separate container from other sites I need on the same domain

!!info

@Bob it's a good idea to update it anyway :P

@Bob I awoke on Fri, 12 Jan 2018 14:26:34 GMT (that's about 43 seconds ago), got invoked 1 times

@allquixotic I remember it being a pain last time

...I don't

Will people ever snap out of smartphones or we are all doomed?

I tried both 3.5.3 and 3.5.0 back then. Neither worked.

3.4.0 worked fine

unless they've maintained stuff poorly (which could happen), the latest stable FF + the latest stable geckodriver + the latest stable Selenium-Java should all work together

oh you're using selenium server?

I need to take a closer look at what you're doing :P

@allquixotic ya, standalone server

webdriver (bot) connects to standalone selenium server via webdriverio

@Bob is that because you're hitting selenium-server via JS?

got it

standalone server launches geckodriver and manages it

@allquixotic yea I suspect webdriverio is incompatible with whatever hub/grid thing the newer ones do

can't remember but I think I tried updating webdriverio and it either didn't fix the issue or it broke more

... probably a simpler architecture to do the selenium client in-process, which would be perfectly doable if you were using a more supported binding like selenium/java

anyway, selenium version isn't hugely important and I'd rather not rewrite the script right now

lol ok

@allquixotic yea, possibly. can't remember why I chose node/webdriverio

might do a dotnetcore one :D

as a fun exercise I might rewrite it in another language that has an officially supported Selenium port

maybe a language that doesn't have page table bloat as a feature

C++? :S

@allquixotic well, there's an official selenium binding for nodejs iirc, but webdriverio provided some nice abstractions

especially around clicking and waiting, and script execution

the syntax in raw selenium is horrendous

@allquixotic dotnet! :D

@Bob the bindings must be pretty bad for Node; I've never felt raw selenium in Java was "horrendous"

@allquixotic automatetheplanet.com/webdriver-dotnetcore2

but where that did feel a little inconvenient I would use something like Selenide

I might give it a shot if I'm bored sometime

the big three official languages seem to be Java, C#, and Ruby, since they all get simultaneous releases. the JS official bindings haven't been updated since October, but Java C# and Ruby got updates in December

Python might also be an official supported language but it got an update in January for a few bugs that don't appear to be shared with the other three bindings

anyway, the important bits (firefox) are updated now

nothing else touches the internet

@Bob curious, is there any particular reason why a long-running Java process will tend to scatter across your virtual address space, but dotnetcore / mono doesn't? is it a design issue with the GC or something?

@allquixotic apt update and apt upgrade completed too

or the heap allocator I guess, not the GC

you're probably good to restart now

and then run the ~/dev/run-headless.sh script after

@allquixotic I'm not sure, tbh. But I've noticed it in two places on two OSes so far.

I don't really have long running dotnetcore scripts anywhere so idk if it does that too

it's not even a memory leak, right? if it were a memory leak, the mapped memory would bloat

it's just the process using loads of different addresses within the VAS

almost like it picks them at random

@allquixotic That's the other thing, Java expects (?) you to set Xms and Xmx

I've never had to do heap size tuning for .NET

I don't think that's even possible

@Bob same

I don't think even Ruby has that problem and people run that in production all the time

to practice for my Xamarin/C# Spotify project I might do the C# impl of the bot loader script; I also need to learn NuGet and see if there's a "better NuGet" like Gradle is a simpler/easier Maven

@allquixotic NuGet is the better NuGet :P

:S

@allquixotic I just pmap'd the selenium process (at 6 GB right now)

sean@cavil:~$ pmap 7952
7952:   java -Xmx2g -Dwebdriver.gecko.driver=/home/sean/dev/firefox/geckodriver-launcher -jar selenium-server-standalone-3.4.0.jar -timeout 0 -browserTimeout 31536000
0000000000400000      4K r-x-- java
0000000000600000      4K r---- java
0000000000601000      4K rw--- java
0000000000b61000    132K rw---   [ anon ]
0000000080000000 1376256K rw---   [ anon ]
00000000d4000000  22016K -----   [ anon ]
00000000d5580000 688128K rw---   [ anon ]
00000000ff580000  10752K -----   [ anon ]

it's mapped a couple of 1 GB chunks for who knows why

and that's with Xmx2g

lol, -Xmx apparently doesn't do anything to limit VAS consumption

lol.

@allquixotic I think I might have looked this up before... stackoverflow.com/a/28935176/1030702

fail

so yea it's effectively a memleak in the jvm

umm did they fix it in Java 9 or a Java 8 update? :S

Oracle claims it's fixed

@allquixotic fixed in 8u131

Oracle claims it's fixed in 8u131 but Cavil's running 8u151! :D

awesome. not fixed.

@allquixotic or a similar issue

hmm

I've had a surprising amount of software break with Java 9, hilariously a lot of it due to attempts to parse the version number of Java and not recognizing "9", but I could update Cavil to use Java 9...

meh, can always go back

@allquixotic Apparently if you want to limit heap size in .NET (not Core) you need to use the hosting API

@allquixotic as long as selenium standalone 3.4.0 works

@Bob I don't need to limit the heap size if it isn't growing unbounded! :D

lol

@allquixotic the best part is the java one is somehow growing past the limit I set -_-

hmm I have to update my TS3 servers too (the ones we visit once in a blue moon other than Pathfinder)

@Bob in resident or virt?

@allquixotic virt

@allquixotic bugs.openjdk.java.net/browse/JDK-8164293

> after sweeper memory leak fix, i observe one class module leak. yet to figure out what is causing it.

> i strongly assume there is no leak in mtClass, only steady growth of OopMapCache. This should stabilize after a while. waiting for sometime to confirm this.

downgrade to an unsupported version; obliterate performance; or just slow down the problem. wonderful

???

bugs.openjdk.java.net/browse/JDK-8187763

> It leaks on java1.8.0.91 but doesn't on java 1.8.0.31 and jdk1.7.

> According to report memory leak observed on 8u91 and not on 8u131

uhhhhhh

is it just me or did the guy who commented and closed it misread 31 as 131?

o_O

hi AU :P

@allquixotic The bug reports seem to be largely talking about resident memory... bleh

@allquixotic that ... is faaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaar too much vegemite

oh it's separate from the ration :P