Tavern on the Meta

Journeyman Geek

08:23

→ 13 messages moved to Chimney

Journeyman Geek

08:36

@GetAnswerWizard was actually pretty fast, but badly configured torrent client on mobile ;p

Glorfindel

09:27

@SonictheWizardWerehog it fixes all broken images in a post (if it's able to). I'd be very interested in an example where it didn't...

Sonic the Wizard Werehog

@Glorfindel Here's an example

Glorfindel

What it should do better is taking into account the entire Q&A thread and fix all images at once. At least on sites where I have edit privileges.

@SonictheWizardWerehog ah, of course. The other images aren't available, not even via the Wayback Machine.

Glorfindel

10:02

... for a moment I thought my browser was hacked.

Then I realized which day it is will be tomorrow.

Shadow Wizard

10:24

heh

9

Q: How to disable "time travel" entirely?

Every time I open a question I get this oh so funny 90s feel with scrollers and whatnots. In every question I need to click it off. Is this an attempt to keep people from using the site for a few days or how do I tell you "please never ever do anything like this if you want me to keep contributi...

While I love fun, I admit this is annoying even for me, and the fact we can't disable it is really really bad.

@Shog9 any chance to add the option to disable this like one can choose "I hate hats" during winterbash? ^

Sonic the Wizard Werehog

Can a mod really quickly make back-to-the-future a synonym of time-travel-mode?

Catija

It's a bug. The back to the future button is supposed to disable it everywhere.

Journeyman Geek

Oh dear god

@Catija evilest April fool's day joke ever

Tinkeringbell

Honestly? We're going to have to mod meta-fires because of an April Fool's joke now? :)

Journeyman Geek

Lucky for me I have two days off coming up

Tinkeringbell

10:37

Well, you have fun then! 😛

Catija

And, of course we had to turn it on at some ungodly hour of the day.

Tinkeringbell

@Catija Around lunchtime? Yeah. Really bad choice :P

Shadow Wizard

@Catija so... 6-8 days to fix?

sad laugh

Catija

Looking like they're poking at it, though.

Shadow Wizard

IMO just pull the plug out for now, it's a total show stopper.

The Stack Egg caused a DDoS attack but it was turned off minutes after balpha realized that (or was it Nick?), they didn't let it run while broken.

EKons

10:50

alright... WHO DIDN'T PUT UNICORNS IN THERE

Journeyman Geek

@ShadowWizard probably not balpha :/

@Tinkeringbell on the other hand weekend

Tinkeringbell

@JourneymanGeek Yeah, weekend. You know, when there's supposed to be almost nothing happening :P

Catija

Looks like it's fixed for me on a per/site basis. Can y'all confirm?

Tinkeringbell

@Catija Yes

Shadow Wizard

@Catija not fixed, disabled. Gone.

Sec, let me try new browser...

Tinkeringbell

11:01

That's true, I can't turn it back on

EKons

they should've included an option to delete the cookie...? :P

Shadow Wizard

Well, better than before, but still... not what I expect.

John Dvorak

I've already written a userscript that clicks the time travel icon if the page is FUBAR'd. Are you saying I won't need to work on it anymore?

Shadow Wizard

@JohnDvorak yup

EKons

@Catija um, the pop-up still appears every time even after I X it on one page

Catija

11:04

That may be intentional...

Can check.

We're looking at the stickiness of that anyway.

EKons

yeah the stickiness is too... modern :/

Catija

Weird... I opened a new site and it took a few seconds for the popup to appear.

EKons

yes me too

Meta Bug Wizard

The ~~title word filter~~ time machine "theme" is one of the worst ideas ever implemented on SO.

EKons

but it does eventually appear, and... I've told it "choo!" but it keeps on bugging me...

John Dvorak

11:07

I can't think of anything worse, design-wise at least.

Meta Bug Wizard

well, the title word filter is still worse than this "theme" (or "breakage"), IMO

EKons

also, there's a little surprise if you scroll down enough... yeah, doesn't look great

Meta Bug Wizard

I crafted that "set cookie" js into a userscript and am feeling much better now :)

Catija

The delay is intentional. Adding a cookie to the X in the popup.

EKons

yeah, the owl also did :P

> The delay is intentional.

that really makes you think "wtf is going on here am I hacked"

Shadow Wizard

11:10

@EKons don't click the "x", click the clock icon.

The "x" just close the banner.

EKons

I will :P

but not sure if it works network-wide yet

Shadow Wizard

@EKons should tell it "chu!" instead.

It's oo sensitive.

@EKons oh, it's not and it won't be.

Nothing like that was ever made network wide, e.g. winter bash: you need to disable it on every site.

I'm fine with that, only very few people are active on many sites, so it's a rare corner case. Not worth the time the dev should spend on it.

That said.... some global "I hate fun!" preference in the profile would be... fun. :P

Meta Bug Wizard

last year's yellow duck is a really good design

Shadow Wizard

@MetaBugWizard agreed. But..... the bad design now is part of the prank.

That's how sites looked like back then.

Meta Bug Wizard

principle 1: don't cause breakage

John Dvorak

11:18

This is a circus though :P

Pandya

It should show "Welcome to ~~Stackoverflow~~ Stack Exchange!"

Shadow Wizard

@Pandya nah, the company name is Stack Overflow

Pandya

The network name is Stack Exchange I think.

Shadow Wizard

meh... fix the big bugs first. ;)

John Dvorak

Indeed - Stack Exchange is Stack Overflow to the extent that Google is Alphabet. Anyone heard of Alphabet?

Pandya

11:23

> Stack Overflow is part of the Stack Exchange network

Tinkeringbell

@JohnDvorak I did!

John Dvorak

the company name is Stack Overflow. For some arcane reason.

Catija

Brand recognition.

Shadow Wizard

Also, can't see a way to flag/block the guest book messages.... they're open to spam/offensive crap.

Pandya

Especially on non-technical sites, the network name is preferable rather than company name.

Shadow Wizard

11:24

@JohnDvorak saw an article about it long ago, forgot about it totally, until now. lol

Tinkeringbell

@ShadowWizard I thought they looked like a set of preprogrammed messages showing up

I dunno if it'll show actual messages from other people?

Shadow Wizard

@Tinkeringbell nope. It's real and live.

Oh wait

lol

Maybe just showing own messages.

calling sock

LOL

yeah, all good.

Looks like the devs had some internal fun, collected bunch of those, and to make it appear real it's also adding locally what you write. All good. :D

John Dvorak

By the by, could the timeout length pleeease be added to request quota error message? In case I click a button multiple times thinking it would disappear while trying to help millions of users ASAP some day again?

Shadow of the Shadow's Shadow

@ShadowWizard did you call?

Shadow Wizard

@ShadowoftheShadow'sShadow yes

kicking sock away

Pandya

11:31

Lol!

Journeyman Geek

...

Pandya

I think the only way to came out of time travel is Install Netscape 3.0! Haha

John Dvorak

Can I even download Netscape Navigator?

Sonic the Wizard Werehog

0

Q: Why was the decision made to make the 2019 April Fool's joke enabled by default?

So far, as far as I remember, every April Fool's prank that SE has made has been non-intrusive, only really showing up as a minor sidebar while not affecting the site's design much. However, this year, the Time Travel Mode prank made drastic, intrusive changes to a site's design, which is not li...

support april-fools time-travel-mode

John Dvorak

@Pandya Ironically, that might actually work. Because NS3 doesn't support Javascript. Or at least not jQuery.

Language experts here, does this count as a disc jockey? diamondback.com/media/catalog/product/cache/1/image/1800x/…

Journeyman Geek

11:37

@JohnDvorak Yes

rene

@JohnDvorak ask on retrocomputing.se ....

Get Answer Wizard

reminded me of GeoCities <3

Sonic the Wizard Werehog

The only great thing I can say about Time Travel Mode: it loads perfectly in IE11.

Pandya

Built in Notepad!

rene

It is honestly awesome

Get Answer Wizard

11:47

Too bad those who were born in 2000s can't really appreciate this...

John Dvorak

Neither do those born before 2000

Olivia

!!/watch matteorusso\W*(?:dot\W*)?business\W*(?:at\W*)gmail\W*(?:dot\W*)?com

SmokeDetector

@Olivia That pattern looks like it's already caught by Bad keyword with email in body and Bad keyword with email in answer; append -force if you really want to do that.

Olivia

huh?

!!/test [email protected]

SmokeDetector

> Would not be caught as a post, title or username.

Olivia

11:48

Noice.

!!/watch-force matteorusso\W*(?:dot\W*)?business\W*(?:at\W*)gmail\W*(?:dot\W*)?com

John Dvorak

Your test doesn't match your watch

Olivia

crap xd

!!/unwatch matteorusso\W*(?:dot\W*)?business\W*(?:at\W*)gmail\W*(?:dot\W*)?com

EKons

um... the . doesn't get caught...

neither does the @ and the other .

Olivia

!!/watch-force matteorusso\W*(?:dot\W*)?business\W*(?:at\W*)?gmail\W*(?:dot\W*)?com

@EKons \W* catches both

John Dvorak

?: doesn't invoke predefined character classes. It's just a grouping construct.

Olivia

11:53

regex101.com/r/vDK2AT/1

Pandya

For April fool chat.SE should show "Under Construction, try IRC until chat feature is released" ! Lol

Catija

OK. Toggle is fixed. ... and... so much for my last two hours of sleep tonight. :sigh:

Tinkeringbell

@Catija I'm sorry 😬

This sounds like the worst April Fool's for everyone ever.

3

rene

We fooled your sleep ...

Olivia

@Tinkeringbell Seconded

rene

12:07

@Tinkeringbell and that by itself makes it awesome

John Dvorak

No sleep is deserved by those dare to bring the 90s back. Those who would give us up. Those who would let us down. Those who would run around and desert us.

Get Answer Wizard

Tony the Pony. He comes.

rene

@JohnDvorak lol

John Dvorak

Zalgo was a good meme.

Tinkeringbell

@rene Hahaha. This 90's kid likes the 90's design (even though I never once visited the internet in the 90's). But it having a bug like that, where it's not easily turned off, is understandably making some people a little bit more angry than they should be on a sunny Sunday afternoon ;)

John Dvorak

12:08

Or rather, Bobince's answer was.

rene

→ 6 messages moved to Trashcan

solved, let's move on

Sonic the Wizard Werehog

I remember following an old HTML for Kids tutorial that was based on HTML 3.2

Get Answer Wizard

I don't know if HTML has versioning before HTML5, because at that time, everything with X was awesome, e.g. XHTML

John Dvorak

Was it on w3schools, perchance? I bet it's still there...

rene

Wait. There is a newer version of HTML?

John Dvorak

12:14

@GetAnswerWizard well, except if the x was in cx

Sonic the Wizard Werehog

@JohnDvorak No; but it remained online until around 2016 or so (when it was taken down and replaced with a modern HTML5 version)

@GetAnswerWizard Well, the first two versions were called HTML and HTML+; it wasn't until later that they used versions, with 2.0, 3.2, then 4.01

Get Answer Wizard

(note to myself: beware of mods fooling around by editing "worst" into "best" later ;)

Sonic the Wizard Werehog

Reminder: when you have tens of millions of users, anything you do or change make will make some subset of those users angry.

Tweeted by Nick_Craver on March 31, 2019 at 12:18 PM

Really?

Tinkeringbell

Yes.

Though doing it in a buggy way makes them even angrier ;)

Magisch

It's true. You notice that effect even if you have only 1000

Sonic the Wizard Werehog

12:20

That's their response? That "we don't have to do anything because the majority is OK with it"? What happened to being inclusive to everyone?

Magisch

Maybe it's just idle musings of nick

John Dvorak

I take offense in that tweet's insinuation that what SE did this year was perfectly OK

Magisch

I dont know you can read that into that

rene

Oh , just cool down will you @SonictheWizardWerehog don't make a drama out of a joke. have a laugh.

Tinkeringbell

I don't see a 'we don't have to do anything' in there. Just an acknowledgement that it made people angry. But if y'all want that fight with Nick, take it to Twitter so I don't have to mod it.

Sonic the Wizard Werehog

12:22

@rene No, it just goes to show their general thinking.

Magisch

nick is a sysadmin, he could just as well be talking about some of the SEDE bugfixes that removed bugs people relied on for some stuff

Sonic the Wizard Werehog

@Tinkeringbell That's exactly what I did.

Get Answer Wizard

grabs corn pop

Sonic the Wizard Werehog

@GetAnswerWizard Not sure if a single piece of cereal would be enough...

John Dvorak

Fine, fine, fine. I'll suspend my disbelief in incredible coincidences :P

Olivia

12:23

14

Q: What testing is done on 'April Fool' pranks?

So, we have just had an 'April Fools' jape that activated on 31st March for many people wasn't capable of being disabled site-wide (or even better account-wide) presented a design (colours, moving marquee, random sparkles) that for many people renders the site physically unusable (admittedly th...

discussion april-fools time-travel-mode

6

Q: Please don't repeat such unprofessional "full design" April's Fools in the years to come

TL,DR: this year's April's fools could be a source of confusion, or worse, in certain workplaces. Please stick to less intrusive stuff. Very related: Why was the decision made to make the 2019 April Fool's joke immediately activate?, but this asks about the "why", and is not a feature request ...

feature-request april-fools time-travel-mode

4

Q: Is it intentional that the April Fool's time travel feature has to be deactivated on a per-site basis?

It seems to me that the time travel feature has to be deactivated on a per-site basis. I've tried this in Edge and Chrome. Is this intentional? It's fine for people using one or two sites, but gives a poor experience if you are using a lot of sites, or just browsing through Hot Network Questions...

discussion april-fools time-travel-mode

Not to forget the bug

Last year was fine. slightly annoying, but it was at least funny

When the update hit, I thought I had gotten some nasty browser plugin installed, followed by the inability to disable it and the issue that it essentially renders the site useless. How is that funny?

Tinkeringbell

I know. All of them are legitimate questions. All of them deserve to be talked about. All of them also deserve that we talk about them in a calm, rational way.

If you don't find it funny because it was just too much this year that's good feedback...

If there are other things you'd rather see on April Fool's that's also good to mention.

Olivia

How about something that doesn't make the site look like it has been hacked, and there's a movie-style panic in the HQ and lots of screens flashing?

rene

@SonictheWizardWerehog no, it doesn't. It shows they do appreciate a good joke, and even better if it can't be switched off. And for that I praise them. You don't have to like the joke. Blame them again when they ignore you on April 3th. Today is not a good context.

Get Answer Wizard

FWIW, and this is my honest opinion: aside of the bug, I love it. It tells me that they're now confident fiddling with site design thanks to the unified theme.

2

Tinkeringbell

@Olivia But that's the joke!!! 😛 I would totally personally say it was pretty obvious this was just a joke. But this did create a bit too much of a panic, I agree :)

rene

12:28

@GetAnswerWizard I posted something to that extent on MSO ...

Olivia

@Tinkeringbell The joke was attempting to make the site or browser of the user look hacked?

John Dvorak

I'd be fine with the joke if I could disable it network-wide with just a few clicks (rather than several minutes of furious coding that got me locked out of the entire network because a button I thought would disappear didn't).

Tinkeringbell

meta.stackexchange.com/q/326018/369802 and meta.stackexchange.com/q/326023/369802 are nice calm Q&A's to look at 😀

@Olivia Probably not. But then again, I have little or no experience with hacking or being hacked, and so didn't immediately think of it as that.

John Dvorak

Having to kill it on each site manually is ... bearable, but pretty much not fun.

Tinkeringbell

@JohnDvorak Well, that's probably something you can request for April Fools to come...

John Dvorak

12:31

I'd hope it'd go without saying. But noted to make a meta post that says it.

Olivia

Nasty stuff being added to a website is one of those things... Mainly through search plugins, but also redirecting the start page to NSFW stuff (haven't had either personally, but I know people who've gotten those types of viruses). I actually tested a browser plugin I thought was spam (which I was also correct about - it attempted to post an answer with a link to itself and also made the textboxes look weird). The textboxes ended up with a gaussian noise-like appearance.
Marquees and bright colors or other forms of distortion where there usually aren't any is usually never a good sign.

Tinkeringbell

I'll keep it in mind ;)

Get Answer Wizard

If you think you got hacked... well, so you know, that's how GeoCities pages were built.

just do google image on it

(afk for dinnering)

Olivia

And to be quite honest, having burned myself on various problems, I'm extremely careful to avoid them. That's why I have an HTTPS enforcer, a plugin that loads jQuery and other CDN JS stuff locally, along with 3 ad blockers and an additional tracker blocker, and JS blocking, to avoid stuff like XSS exploits on MSO

Some stuff appears where you don't expect it, and can be either harmless, annoying, or worst case scenario plain out damaging

John Dvorak

Serious question, willl we have managed to kill some photosensitive epileptics during this year's April Fools' prank?

Olivia

12:39

And a bug that makes it impossible to disable doesn't exactly inspire a "this is by design"-kinda attitude...

@JohnDvorak I wondered about that too:

in Dawg's Waffle Haus on Stack Overflow Chat, 2 hours ago, by Erik A

It isn't. That's just something people say with colorful things, as long it's not bright and high-frequency flashing that doesn't really happen

Journeyman Geek

@JohnDvorak hopefully no

John Dvorak

The thing is, SE has no way to ensure they haven't. Those guys aren't going to post an angry rant from behind the grave.

No matter how much I'd love to see an angry rant from behind a grave.

Tinkeringbell

@JohnDvorak I think if one of my loved ones died due to that, I'd at least shoot the site a message 😉

Not really from 'behind the grave', but they'll find out eventually.

Americans make lawsuits for everything, right? :P

John Dvorak

Not the dead ones, those don't sue :P

You're right though, SE would probably get the news if they did maim someone.

Shadow Wizard

You win some, you lose some.

Rob

13:08

Microsoft put out a memo a few days ago banning pranks tomorrow, April 1st.

Shadow Wizard

:(

balpha

what if that memo was just an early april fools joke?

Shadow Wizard

hehehe

John Dvorak

Then they should make another prank as well, turning the former into a meta-prank

Shadow Wizard

I'd love a prank where the CEO sends a "YOU ARE FIRED!1!!" email to all employees. :P

glancing at @Joel

Tinkeringbell

13:16

@ShadowWizard Prepare for lawsuits, having to pay severance pays etc...

Shadow Wizard

lol

Luckily for everyone, I'm not CEO. :D

balpha

@JohnDvorak They should make another memo that says the earlier memo was an early april fools prank. Then either 1) the new memo is real, in which case april fools pranks are okay, therefore goto 2, or 2) the new memo is a prank, in which case the old memo holds, in which case the new prank violates policy, therefore goto 1)

Shadow Wizard

endless loop detected

Shadow Wizard

13:41

Tried to install Netscape 3...

:-(

Tinkeringbell

Time to drag up a VM :P

Shadow Wizard

lol

Journeyman Geek

I had an OS/2 vm with netscape and opera on my old PC ._.

@ShadowWizard Mosaic will work on modern linux ;p

Shadow Wizard

:D

Journeyman Geek

59

A: Why can't old PCs view modern sites?

Simply, a lot of modern HTML features we take for granted did not exist back in those days. There are attempts to build 'modern' browsers for older OSes - classilla, for example. You can easily load up linux onto a suitable PII or PIII system and still be able to handle a basic modern webpage. ...

Mgetz

13:55

@JourneymanGeek I remember being specifically told not to go to sony.com back in the 90s because it used a TON of images and wouldn't load. Not because the computer couldn't handle it (it couldn't) but because our dial up was too slow. So I tried at school and it was still broken as all get out.

Journeyman Geek

@Mgetz it wasn't the connection in this case ;p

Mgetz

oh I know Ars Technica did a similar article for the death of IE6

Journeyman Geek

.... actually ....

@ShadowWizard I wonder if I could throw this on a VM for maximum luls

Mgetz

what amazes me is I have clients that insist on using IE11

Journeyman Geek

utilu.com/IECollection

@Mgetz ._.

I THINK we're on 11 now. Might have been on 10 before

which is doubly fun cause I use a bit of a niche browser by preference

Mgetz

14:04

@JourneymanGeek I refuse to support IE11

thus far no client has called me on it

Journeyman Geek

@Mgetz we can't upgrade our machines, also they're off the internet.

Mgetz

mostly because I can say with certainty that MS would prefer we didn't

Journeyman Geek

MS's biggest competion is.. their past selves

so that makes sense

Shadow Wizard

@JourneymanGeek someone appears to already do this...

18

Q: Does not work on Netscape Navigator 3.0

Finally, after 20 years of suffering, StackOverflow now claims Yet all I see is NOTHING! Have all those years been wasted? Have I endured 20 years of ``error: insufficient memory'', browser segmentation faults, and JavaScript exceptions just for this? They lied to me! All I want are so...

bug fun april-fools time-travel-mode

That's why I wanted to install, to reproduce... ;)

Journeyman Geek

looks like windows 98

Shadow Wizard

14:07

I also didn't find what response Netscape 3 expected from the server, was going to post a serious answer with that info. :(

Journeyman Geek

Presumably on VMware, its a pain in the rear getting it to work on vbox

Shadow Wizard

Or is it just "wrong" HTML, not recognized by the browser?

Mgetz

I had to go back to the future, that was hurting my eyes

@ShadowWizard it might not recognize a doctype

HTML2 didn't have them I don't think

Shadow Wizard

@JourneymanGeek 11 is the last. If one says IE 12 they probably mean Edge.

(which is wrong, but meh)

Journeyman Geek

@ShadowWizard we were on whatever came standard with windows 7

Shadow Wizard

14:10

@JourneymanGeek oh... think 11, yeah.

And 10 pretty sure came with Edge already.

Mgetz

@JourneymanGeek IE8 then

Shadow Wizard

So confusing lol

Journeyman Geek

THen we're possibly on 9

Mgetz

although that's very very unsupported now

As MS only supports the latest IE for each OS

Journeyman Geek

well a good chunk of our tech stack is...

Mgetz

14:11

Industrial tooling?

Journeyman Geek

soon to be unsupported ;)

We run servers with a name rhyming with a certain doomed british vessel

Mgetz

Kinda surprised MS doesn't make an OS just for industrial tools, a very very stripped down NT stack

Journeyman Geek

@Mgetz actually...

Mgetz

@JourneymanGeek still supported by HPUX!

Journeyman Geek

windows embedded and such..

@Mgetz that's our OS

rene

14:13

@ShadowWizard try the 16bits version?

Mgetz

@JourneymanGeek AFAIK HPUX will be supported for quite awhile

at least on Itanium

Journeyman Geek

@Mgetz Itanium's EOL in... 2022?

Mgetz

that said your power efficiency is probably crap

@JourneymanGeek as far as new production of chips yes

but HPUX will probably be supported on it until 2030

Journeyman Geek

@Mgetz and "new" chips are a 2017 design at best?

Mgetz

@JourneymanGeek 2015 IIRC

Journeyman Geek

14:15

I'll be out of here by then ;)

Not sure where, but out

Mgetz

The process node is massive in comparison and intel has no intention to move

They could cut the power usage in half by moving it to 14nm

but they won't

Journeyman Geek

Kittson is 2017 - that's the last itanium generation

also I wonder how it compares, core for core with a modern x86'

Mgetz

last released, the processors were taped out ages ago

@JourneymanGeek depends on compiler

Journeyman Geek

ah

ehm...

java

Mgetz

You'd be better off buying Arm A57s

Java never had great support for Itanium

Journeyman Geek

14:18

I'm just a lowly level 1 support engineer ;)

Mgetz

mostly because Itanium really needs ahead of time compilation support

EKons

@Olivia but you're preceding them with dot and at...?

so it will capture [email protected], for example

!!/test [email protected]

SmokeDetector

> Potentially bad keyword in body, potentially bad keyword in title, potentially bad keyword in username
----------
Potentially bad keyword in body - Position 0-38: [email protected]
Potentially bad keyword in title - Position 0-38: [email protected]
Potentially bad keyword in username - Position 0-38: [email protected]

EKons

!!/test [email protected]

SmokeDetector

> Potentially bad keyword in body, potentially bad keyword in title, potentially bad keyword in username
----------
Potentially bad keyword in body - Position 0-30: [email protected]
Potentially bad keyword in title - Position 0-30: [email protected]
Potentially bad keyword in username - Position 0-30: [email protected]

EKons

14:20

hm...

looks like something was changed while I was away

Journeyman Geek

@Mgetz I'm not exactly a fan of this place, but employment and money are... well something I am a fan of ;)

I'd be happier if we were running something a bit more uptodate and a little less unusual

I can't even use backspace on our shell

Mgetz

@JourneymanGeek fair, I'm actually very curious to see what will compete with Xeon in the server

Journeyman Geek

@Mgetz epyx and her chinese cousins seem obvious

Mgetz

@JourneymanGeek I've found the more you're off mainstream the more expensive it gets

@JourneymanGeek Possibly RISC-V or OpenPOWER

Journeyman Geek

@Mgetz naw, I'm basically getting paid warm body rates

@Mgetz ironically it needs to be cheap and cheerful and scale well

POWER was never cheap and cheerful

RISC-V seems interesting

Mgetz

14:25

@JourneymanGeek They seem to be working on that to an extent but you're right, I think RISC-V stands a better chance of doing the "Thousands of cheap cores" bit

Journeyman Geek

yup

and the fact that its starting as an open ISA

Mgetz

Oh I don't think that's necessary

I think the fact that they aren't charging a ton for a license helps

Journeyman Geek

(also, mips is deadish, via's x86 is... kinda an odd limbo)

Mgetz

so google etc can experiment cheaply

Journeyman Geek

yup

Mgetz

14:27

I've expected Apple to go their own way for awhile now

Journeyman Geek

and WD's already swapping arm controllers with those

I don't.

Iphone aside, there's something to be said for commodity components.

Mgetz

you think apple cares?

Journeyman Geek

@Mgetz its more profitable.

Mgetz

There is a reason the xcode compiler is closed source

John Dvorak

speaking of MIPS, I really like the idea of delayed branches. Too bad there's usually nothing to actually do while waiting for the branch to happen.

Journeyman Geek

14:28

I think they'd keep the arm option open to strongarm better deals from intel...

Mgetz

@JourneymanGeek apple already makes their own processors I don't see them sticking with intel if they think they can increase margins

Journeyman Geek

@Mgetz they design them, sure, but intel basically still has a big edge on fabrication over everyone else

Mgetz

@JohnDvorak Raymond Chen actually did an article on why delay slots actually really really mess things up at a silicon level

Journeyman Geek

and apple's a bit of a halo customer

John Dvorak

@Mgetz link?

AFAIK, modern CPUs' pipeline controllers are a hot mess no matter how you look at it, having to co-ordinate three instructions at once or balancing 8 logical registers between 32 physical ones...

Mgetz

14:32

@JohnDvorak devblogs.microsoft.com/oldnewthing/20180416-00/?p=98515

good place to start

John Dvorak

Thanks

Mgetz

@JohnDvorak trust me that's easy compared to dealing with interrupts at a point that's not friendly to interrupts

having a base ISA that's interrupt friendly makes the rest of that stuff easy

John Dvorak

Can you just ignore interrupts while having branch delay stuff in the pipeline?

Mgetz

@JohnDvorak no because that would allow denial of service

John Dvorak

Disallow overlapping branch delays / only if an interrupt is pending?

Mgetz

14:35

@JohnDvorak How is user mode code supposed to know that?

John Dvorak

I would expect the CPU hardware to know about interrupts

Mgetz

inherently interrupts aren't always pending and some are non-maskable

@JohnDvorak it does, but that means you have an instruction that may not get executed. So... why have it?

instead just block the pipeline for the branch

John Dvorak

I suppose modern CPU hardware can generate branch-delay execution plans when none are present in the ISA anyways?

Mgetz

the main TL;DR: don't expose your pipeline via your ISA

@JohnDvorak we call that speculative execution

John Dvorak

I mean, I imagine if I have DIV BX, CX; JMP loop$1, the CPU can jump first and then start waiting for an ALU to free up.

(Although, division can trap, so that would be an issue...)

Mgetz

14:42

@JohnDvorak depends on the ISA, on x86 it can because there are no memory interactions. Ditto ARM and most others

most compilers would MUL by the reciprocal anyway

pretty much avoiding any possibility of using DIV at all costs

John Dvorak

I'm still a bit puzzled how you can perform integer division via modular multiplication.

It's easy if you know there's no remainder, but you don't.

Mgetz

@JohnDvorak if either part is constant it's easy?

John Dvorak

Is it?

Mgetz

@JohnDvorak godbolt.org/z/fCD5LE

John Dvorak

return 3 / num produced IDIV just as I expected

Mgetz

14:45

in most cases the divisor is a known quantity

@JohnDvorak yes but the vast majority of cases don't do that

It's worth noting there are ISAs where that kind of Multiplication or division IS async btw

devblogs.microsoft.com/oldnewthing/20180404-00/?p=98435

Magisch

I

John Dvorak

> If you try to read a value from HI or LO, you must wait two instructions before performing any operation that writes to HI or LO. Otherwise, the reads will produce garbage

Excuse me???

Mgetz

> (Tricky rule number two was removed in the R8000. On the R8000, if you perform a multiplication or division or MTxx too soon after a MFxx, the processor will stall until the danger window has passed.)

John Dvorak

Still, how does a write to a register affect the read two ticks before it?

The one that's already happened?

rene

@JohnDvorak common practice if it involves IO.

Mgetz

14:56

because the Multiplication and Division hardware is clocked differently than the rest of the chip

and runs in independent pipelines

basically it's a co-processor

John Dvorak

Still it implies that HI:LO reads happen out of order...

Mgetz

@JohnDvorak possible, the point is that you shouldn't mess with them until the processor manual says it's OK to

either way it was exposing a pipeline feature to the ISA

John Dvorak

What happens if I read from HI:LO, and then use that value to corrupt the read of that value?

Mgetz

no idea!

John Dvorak

I'd expect it stalls the next instruction, and fail to fail the read

otherwise you have a causal loop. Those are not nice.

Mgetz

14:59

I'd expect that prior to the R8000 it would completely corrupt the entire calculation

after the R8000 it would stall

Transcript for