Retro Challenge 2017/10 Zork First Stream

At 7pm UK time on Sunday (yesterday), I did the first Zork live stream.  This post is going to look back at how this went, from a technical point of view, from my point of view as a new YouTube streamer, and from the point of view of a non-adventure gamer playing Zork for the first time.

There was two aspects to the hardware.  The RC2014 running Zork was almost completely stand-alone.  It had it’s own keyboard and monitor, and the only connection to the outside world was via the ESP8266 module.  The ESP connected to the internet over Wifi, and streamed everything that was sent to the monitor via web sockets.  The Tx line from the ESP also went to a laptop so that I could monitor how many connections were open.

The laptop was also running Wirecast, and streaming it’s webcam and microphone to a live (live-ish.  About a 10 second delay) YouTube feed.  I could then talk, mumble, gesture, or use the chat feature, as well as read the chat messages as they came in.

Overall, this worked reasonably well – however there were a few things that could have been better.  About 20 minutes in, there were a couple of people complaining that they couldn’t connect to the text stream from the ESP.  The 3 people still connected were ok and getting updates, but nobody else could connect.  After a reboot everyone could connect and I was streaming to an audience of 5.  There was a report of odd characters being sent, although I think this was only when I used the delete key, which couldn’t be displayed properly in the live feed.  Also, there was still the issue with some lines being truncated.  If anybody viewing has kept a transcript, it will be interesting to compare with my transcript to see if they are both truncating at the same point (indicating a dropout between the web socket server and the ESP), or if they are different (indicating a dropout between the web socket server and the viewer)

I felt a bit awkward on YouTube.  It was kind of odd, and I couldn’t quite decide if I should just be playing Zork on my own and ignoring everyone looking over my shoulder – or if I should be ‘performing’ for the audience and involving them more.  As someone that isn’t that comfortable in front of the camera, and not good at text adventure games, I am beginning to question my life choices that lead to me narrating my fumbling through a troll filled underworld live on YouTube :-/

Although I had loaded up Zork a few times in the past, I had never gone further than randomly typing in directions, or going in to the house.  So I really was treading new ground here.  One of the things that surprised me for a 40 year old game is just how large the place is, and how much is going on.  I didn’t realise that there would be other characters that moved around, stole things from you and dropped things in different rooms.

The first game didn’t last too long, after picking a fight with the thief left me injured and a troll finished me off.  So I restarted the game, and decided to take a different approach in attacking the thief – but was surprised to find he wasn’t where he was in the previous game.

I did find myself questioning my motives here though.  Basically, I broke in to somebodys house, took all of their possessions, and tried to kill the first person I met.  I should be asking myself who the bad guy is in this game!

I had downloaded a map from the internet to make things a little bit easier for me.  It saved me from trying to go in directions that didn’t exist, and meant that I didn’t have to map things out for myself as I went along.  However it didn’t mean that I could just get to where I thought I wanted to go.  For example, once I went downstairs from the house, I couldn’t get back up again.  Well, one way up was up an unclimbable slope, one was up through the trap door that was locked behind me (probably by the house owner after I stole his lunch), and the stairs to the kitchen could only be climbed if I dropped most of what I was carrying.

Eventually I ended up at the Dam, and there was a control panel with a bolt and a green bubble on it.  In the Maintenance Room I found a wrench.  Also a tube of magical gunk.  So, surely these things were all related, and something good would happen if I could remove the bolt.  But no matter what combination of words I used, it just wouldn’t do anything.

I’d been playing for an hour by this time, so decided to call it a day for now.  I’ll resume again soon, although I can’t promise that I won’t have looked up how to remove the bolt before I try again :)

A transcript of my adventure can is below

The YouTube stream can be found here https://youtu.be/Gdjq2Rb9y64 But, really, it’s probably not going to be one of those thing I expect anybody to sit through!

 

Retro Challenge 2017/10 – Live Zork Streaming Soon

Firstly, apologies that not much has happened recently.  I’ve been under the weather for the last few days, so only made minimal progress.  However, feeling a bit better, so Tonight I will try streaming Zork!

I am planning to go live at 7PM UK time tonight for around an hour (If you’re not in the UK, see this countdown timer).  If you missed it, I’m sorry – but I hope to have a transcript up soon, and will probably do it again shortly.

There will be a tweet beforehand with links to connect to, so if you don’t follow @ZXSpectROM on Twitter, click here to view

Essentially, there will be two aspects to the streaming.

  • Primarily, there will be a RC2014, running CP/M, with Zork on it.  This machine will have the ESP8266 Wifi bridge, and everything that is sent to the screen will also be sent to http://tzapu.github.io/WebSocketSerialMonitor/ via web sockets.  This machine will also be using one of the old parallel 80’s keyboards I investigated in the last Retro Challenge (It’s important to point this out so you’ll cut me some slack with all the inevitable typos that I know I’ll make with this keyboard layout).  This will be the live stream of Zork, which is essentially the goal of this Retro Challenge.
  • A live stream on YouTube will also take place.  Well, live-ish.  Apparently there’s a 10 second delay, so that might be a bit odd.  However, the main reason for this is to provide a feedback loop for those watching along at home via the YouTube Chat function.  I’ll have a laptop set up next to the RC2014 so I can see any comments as they come in.  So, if I forgot to pick up an object in a room, or you think I should go north instead of east, you can let me know.  It will be streaming video from a webcam too, but don’t expect that to be too exciting.  (Seriously, you’ll be watching a 40 year old ASCII based text adventure game being played by somebody that isn’t good at text adventure games, and the webcam footage will be LESS exciting!)

 

So, that’s the important stuff that you need to know to follow along tonight.

In other news, as I was happy enough with how the prototype PCB seems to be working, I’ve laid out a real board now.  The PCBs are currently being manufactured in China, although it isn’t looking likely that I’ll get them back before the end of this particular Retro Challenge.  But if I do, then expect an update here!  Here’s a sneaky peek at what they should look like though;

 

Retro Challenge 2017/10 Testing

The initial testing of the ESP8266 board with the RC2014 was pretty good, and fundamentally it worked.  However, it wasn’t quite right, and I suspected that the problem was to do with CR or LF.

The code I was using was found here; https://github.com/tzapu/WebSocketSerialMonitor (Thanks Tzapu!).  It uses web sockets, and allows an external web page to connect through to the ESP8266.  So, by going to http://tzapu.github.io/WebSocketSerialMonitor/ and connecting to ws://x.x.x.x:81/ws (where x.x.x.x is my external IP address and a firewall rule is set up to forward port 81 through to the internal address of the ESP) it will display everything the ESP receives on its Rx pin.  Well, almost everything, but not quie everything.

If, for example, I did a directory listing which was 2 full lines and a little bit more on the 3rd line, only the first 2 lines would show up.  If I’m playing Zork, sometimes it would show the whole chunk of text as it came in, other times it would miss the last line.

The code itself is fairly easy to read, although the complicated web sockets stuff is hidden away in libraries.  This is the part of the routine that reads the serial input until it detects a CR (the ‘\n’ part), where upon it then sends the line;

I figured that adding a check for a LF would do the job;

But, sadly, it didn’t.  If anything, it made it worse.  So, instead, I tried just reading a set amount of characters (10 initially) and sending them regardless of a CR or LF, but that didn’t work either.  When I got it to send every character without checking, it worked much much better, although if I sent a very large chunk of text it would miss random bits of it.

So, the problem seemed to be speed related.  The checking for CR was a nice idea by the original author, but it wasn’t something I needed, so time to strip out all the surplus code.  It went from this;

to this;

Much simpler, and it seemed to work perfectly!

All the testing up until this point was done either on my laptop or mobile.  Although they are on the same wifi network as the ESP, it’s connecting out to an external web page before coming back in, so it should be a reasonable test.  But there’s nothing like real people connecting in to really see if it works.  So, I put a shout out on Twitter, and on Saturday evening, 4 willing volunteers sat there watching me do random things on the RC2014.

Feedback was pretty good, and it all seemed to work as expected.  Thanks Thilo for this screenshot;

I messed about with simple programs in BASIC, directory listings, and the opening moves in Zork, with success.  In a moment of madness, I fired up Wordstar.  When the RC2014 is connected to a VT100 terminal, it works great by using escape codes to set where things are on the screen and what colours are used.  However, these escape codes don’t render at all well on the web sockets interface.  Thanks Dave for showing my just how bad it looked!

So, yes, I should have known that Wordstar would be pushing it a bit far.  But otherwise, I was very pleased with the performance.

The biggest problem I had, though, was that of feedback when interacting with the “audience”.  I could type short sentences in to Zork that the viewers would see and understand, even if the Zork engine couldn’t interpret it.  But, of course, there was no live feedback from the viewers.

I’m not sure what the solution to that will be, but it could be as simple as keeping Twitter open on my phone.  Alternatively, things like IRC or Slack could be used (although I don’t want viewers to have to jump through hoops to get connected to me).  Skype might be an option too, although, at the moment, I think it’s more likely I’ll set up a YouTube live stream and use the chat feature in that.

For those wondering about the set up that I’m using, it’s pretty much a stock RC2014 Pro with a Pi Serial Terminal and the ESP8266 prototype I built for this project.  The Pi Zero is connected via a HDMI > VGA adapter, and then to a 17″ monitor.  The keyboard is this Cherry keyboard from the last Retro Challenge.  Note that the key layout is sufficiently different from what I’m used to that typos aren’t too uncommon yet.

The laptop is there to program the ESP, and to monitor it’s output.

I am aiming to have a live run through on Friday, so that gives me a few days to look in to the feedback options.

Retro Challenge 2017/10 Getting Started

So, the idea of streaming the RC2014 to the internet had been rattling around in my head for a while, but I’d never quite worked out how the software side of things would work.  Although the hardware side should be kind of straightforward, with my software skills, I need to find something I can copy & paste.  A couple of days ago I just happened to put the right search terms in to Google and it gave a result that looked like it would do want I wanted.  So, with that, my plan for Retro Challenge took shape!

A couple of years ago I played around with some EPS8266 Wifi Modules, but never really found a use for them.  However, for this, they would be ideal.  The code I had found was a Web Sockets Serial Monitor designed specifically to run on an ESP8266, so I decided to test it out just to check that it did what I expected it would.

Several hours later, it did!  Those hours, however, were spent trying to remember how to program these things.  My desktop machine didn’t have any of the ESP stuff in the Arduino IDE, but luckily, I still have my old laptop that I used a couple of years ago.  Despite getting all the right libraries, it just wouldn’t compile.  So, on to the new laptop, with a fresh Arduino IDE and all the ESP stuff and libraries, and it compiles.  Getting it to upload, however, was another challenge which actually involved burning myself on the USB to serial adapter at one point!  It turns out that of the two I tried, one of them was dead, and the other one spammed my Twitter account ever time it’s turned on (I didn’t know this until checking Twitter later… what!  86 new mentions!).  It also turns out they need more power than the adapter can supply.  And, in order to program them, the ESP needs to be pushed in to the adapter really really really hard!

The programming jig wasn’t really suitable for testing things out, so time to transfer this to a prototype board.  But first, lets work out a circuit!  The programming jig has the necessary support components and minimal connections, so that’s a good start.  The ESP works on 3v3, so I’m going to need a regulator to drop the 5v down.  And the Rx pin is only 3v3 tolerant, so a couple of resistors as a voltage divider will work fine as a level shift here.  The Tx pin won’t be connecting to the RC2014, although even if it was needed, 3v3 is sufficient enough to work.  GPIO0 is used to either put the ESP in to programming mode or run mode.  2 of the other GPIO pins (15 and 2) were tide to ground and 3v3 respectively on the programming jig.  I can’t remember why I did that, but that won’t be needed for this board.

Before hooking up the ESP to the power rails, I checked that the regulator I used (spare one I had lying around) did it’s job.  Yup, 3.299v, close enough!  On with adding the rest of the stuff.

So, from left to right, we’ve got a connector for 3v3 so it can be powered directly if needed (or use 3v3 elsewhere), and an FTDI header for my 3v3 adapter so the ESP can be reprogrammed.  Then there’s a reset button, the 3v3 regulator and caps, then the ESP itself.  There’s a 10k pull up resistor on the reset pin, and a program/run jumper below the ESP.  As yet, the resistor divider for the Rx pin aren’t fitted.  But it should power up and work just as it did in the jig, right?

Ummm… no.  Regardless if I used the power regulator, or fed 3v3 in directly, out of the serial header all I got was gibberish.  It was consistent gibberish, but not something I could make any sense of.  Maybe I damaged the ESP when soldering the wires on?  Hmmmm… maybe I should program another one just in case.  Well, long story short, it needs GPIO15 connected to ground to work!  I can’t see any logic why this would need to be the case, but, whatever, it works now.

I needed to add a rule on the firewall to allow port 81 through to the IP address of the ESP, and after that, I could connect from the outside world and see what my RC2014 was saying!

Things work pretty well, although there’s some kind of issue with carriage returns and/or linefeeds that’s causing some stuff to be dropped, but that’s enough success for one day.  We’ll sort that out tomorrow!

Retro Challenge 2017

Ok, with 5 days of October already gone, I know it’s cutting things a bit fine for putting my entry in for this seasons Retro Challenge – particularly as it started 5 days ago… but, better late than never, right?

So, what have I planned to do that is both retro and challenging? Simple, I’m going to play games! More specifically, I’m going to play Zork, on a CP/M enabled RC2014 that is connected to the internet so it can be streamed live to an audience of anybody that cares to watch me get eaten by a grue.

The first challenge is going to be hooking up the RC2014 to the internet, such that it can stream text. Kind of like Twitch, but in ASCII. Luckily, I’ve got a bunch of ESP8266 modules that have been sitting in a drawer, and I think these are likely to be ideal for converting serial data in to web pages.  I should be able to hook up something on breadboard, and find a bit of code that does the job – although it would be nice if I can get some real PCBs made up for the job.  With just over 3 weeks to go, though, that might be pushing it timewise.

Then there is the challenge of Zork itself.  I often use it as a test game to show people what the RC2014 can do, however, I’ve never done more than a bit of random North, North, East, Look, Get Lamp etc.  I’ve never really played many text adventures either, so this will be a big learning experience for me.  I can’t promise that I won’t use the odd cheat/clue here and there, but if there are actually people watching along online, I’ll happily take suggestions over Twitter.  If things happen to go really well, and I complete it in good time, then I will add Adventure as a stretch goal.