IntelliJ plugin for 6502 assembly language

I’ve been writing some 6502 assembly code in the past year, and have found the developer experience for this language to be lacking some modern conveniences.

In my last blog post, I described the development tools that I used to implement a simple NES game on Linux. I used text editor to write the code, and it couldn’t do much more than syntax highlighting.

I write most of my other code in IntelliJ, so I decided to take a look at what would be required to write a plugin for 6502 assembly support. I managed to together something that mostly works, which I published to the JetBrains Marketplace last weekend.

You can find it by searching “6502 Assembly” in most JetBrains IDE’s.

Features

Firstly, it supports syntax highlighting. I limited the scope of the plugin to ca65 assembler syntax only, since it’s the assembler that I know best.

You can navigate to any label with Ctrl+Click. This will not yet work for other types of symbols, such as constants, macros, and imports.

If the plugin sees a jump or branch statement, and can figure out where the jump goes, then it will show a gutter icon which navigates to the target.

You can find the usages of a label.

In ca65, you can use nested scopes. The plugin shows code folding buttons to collapse these scopes.

You can navigate to a symbol by name using the “Go To Symbol” action.

The plugin allows for commenting and un-commenting blocks of code, though there is no formatter, so indentation sometimes still has to be fixed manually.

It also allows you to rename a label and its usages, which is a great time-saver.

Limitations

One limitation is that the plugin does not fully parse expressions, so it will not detect errors from mis-matched brackets.

Secondly, the plugin does not understand the project structure. This means that if you re-use the same label name in different places in your project, the “Rename” and “Find Usages” function will match all of them at once, because it is not smart enough to follow imports and apply scope rules.

Future improvements

A lot of things can be done with an IDE plugin, but I’m planning to use this initial implementation for a while before attempting to add any big-ticket features.

A reasonable goal might be to have “Hello World” project skeletons for common 6502 systems, and to support launching with some common emulators. I would like to be able to set breakpoints and debug 6502 programs in an IDE, but external debugging interfaces are not commonly available for retro emulators, so this is probably not a reasonable goal.

The code is on GitHub under an MIT license. If you are using this plugin and would like to help improve it, then pull requests are welcome.

Building my first NES game: A retrospective

Last year, I spent a fair amount of time learning how to make games for the Nintendo Entertainment System. The homebrew scene for this system is very much alive, and I was quite proud that I was able to make a simple, working video game which runs on a NES emulator.

So, I present to you: 8-bit Table Tennis.

This blog post just a few notes about the tools and resources that I used for NES development on Linux, plus a few things that I learned along the way.

Development tools

The CPU in the NES is a derivative of the once-ubiquitous MOS 6502, and games for it are mostly written in 6502 assembly. I already do a fair amount of programming, but quickly found that my usual editors, compilers and debuggers were useless for this platform.

The main resources I used were:

To build my code, I used the cc65 toolchain, which includes a cross-assembler. I did most of my testing in the Nestopia emulator, and most of my editing in Gedit. All three of these are available in the official Debian repositories.

Gedit setup

Out of the box, Gedit can’t syntax highlight 6502 assmebly. I found a language spec file for it on the the 6502.org forums, and edited it slightly before using it.

At the time of writing, new language specs can be installed on Debian like this:

sudo cp asm6502.lang /usr/share/gtksourceview-4/language-specs/asm6502.lang && sudo chmod 0644 /usr/share/gtksourceview-4/language-specs/asm6502.lang

It possible to add a console and git support to with gedit, but this is still a long way from a full IDE.

Geany

Geany is a programming text editor, and is also available in the official Debian repositories.

I wasn’t able to get it to recognise 6502 assembly, but it does have good x86 assembly (NASM) support, which is similar enough for it to allow navigation through the source code using labels.

Geany can also set up projects with build scripts (such as a Makefile), which could allow for quicker testing.

FCEUX

FCEUX is a NES emulator which has an in-built debugger, and it runs well under WINE. I was also able to build and run it natively on Linux, but the debugger is only present in the Windows build.

I also briefly experimented with using the FCEUX Lua interface to pause/resume the emulator over a TCP socket for debugging, since that interface is available in the Linux build. I could only pause the emulator between frames, so I decided to abandon this.

Graphics

I created the graphics in the GNU Image Manipulation Program, and manually broke the title page down into tiles.

I also wrote a custom program to convert a 4-colour PNG file into the native NES CHR graphics format. I run this conversion as part of the build process, so that the graphics files can be stored in a modern format.

Things I learned

Programming for the NES is quite simple once the project is up and running, but even simple tasks can be a real hassle. I spent more time than I expected on mundane tasks such as collision detection, and certainly did a poor job of implementing game physics. The 6502 CPU has no built-in way to multiply, divide, or perform floating point operations. The version used for the NES additionally lacks any way to perform a binary-to-decimal conversion, which would have been very useful for displaying the player scores! In hindsight, I would have been able to improve the physics by pre-computing some lookup tables.

On the other hand, I expected it to be difficult to work within 2KiB of RAM, but due to the simplicity of the game, I only used around 25% of it, and had plenty of CHR ROM space leftover as well.

It took me three weekends to make this project, without having ever written a line of 6502 assembly before. I think that development would be a lot faster if I were able to set breakpoints from my code editor, so I will definitely be looking at other debugging emulators before attempting my next 6502 assembly project.

Conclusion

8-bit Table Tennis is available as an iNES ROM on GitHub.

I have only been running it with an emulator, so if any readers of this blog own a flash cartridge for the NES (such as an Everdrive), then please let me know if it works on the real hardware!

Building a tiny Linux gaming PC

I recently put together a small form factor PC to use with my TV, with the aim of building a Linux-based, console-like gaming setup. This mostly went to plan, and is certainly a big upgrade from my previous hardware, which was based on a retired desktop computer.

I am writing a bit about it today, partly because it’s been an interesting project, but also to show a working setup for anybody who is attempting something similar.

Quick reference

The build is based around these components (PC part picker list here):

  • In Win Chopin case
  • AMD Ryzen 5 3400G CPU
  • Gigabyte B450 I AORUS PRO Motherboard
  • Noctua NH-L9a-AM4 Cooler
  • Corsair Vengeance LPX 16 GB (2x8GB) DDR4-3200 RAM
  • Samsung 970 Evo 1 TB

Peripherals:

  • Sony DualShock 4 controller – Works over Bluetooth and can be used as a touch-pad.
  • 8BitDo SF30 Pro controller – Also works over Bluetooth, and can emulate an Xbox controller.
  • Logitech K400R wireless keyboard/mouse

Hardware setup

This is the smallest PC I’ve built with desktop parts, and there is not a lot of spare space in the case. I first put the everything together on my desk confirm that it would POST, then disassembled the case to make some modifications.

Three screws hold in the power supply, two torx screws hold the aluminium shell to the chassis, and two Phillips-head screws and some tabs hold the plastic front panel cover to the chassis.

In that last image, the motherboard almost fills the case, and a power supply has to fit in there as well.

My plan was to route the power cables around the back of the case rather than leaving them in the main cavity. This involved cutting out a square near one of the drive trays for all of the power cables to exit, and another for the ATX power connector to connect to the motherboard. I also removed a metal tab from the power supply, since I had cut out the metal that it was supposed to be screwed to.

There was an existing hole which I could use for the CPU power connector, which was already the correct size.

I then swapped the power supply fan for a Noctua A4x10 PWM so that I could control the fan speed from software. I do not recommend this, since it is unnecessary, and opening a power supply is an electrocution hazard.

Next, I got all of the cables into place, since there is no space to do this after the motherboard is installed

Once the power cables were in the right place, I added the motherboard, then connected the front panel I/O, feeding all of the excess cables to the space behind the case. This is a very tight fit, and I accidentally bent the case and squashed some cables before finally getting the plastic front panel to attach.

After finally lining everything up, I checked that the case closed, and re-attached the shell.

Lastly, I was able to re-fit one of the two drive trays, in case that is ever needed in future.

BIOS setup

This build uses integrated graphics, so the system memory is also being used as video memory. This means that RAM speed is more important than usual.

I enabled XMP to run the RAM at its rated 3200 MHz speed. This is higher than the highest officially supported speed of 2933 MHz for this CPU, but for me that was not an issue.

I set the built-in motherboard LED’s to blue, to match the default colour on the Sony DualShock 4 controller. This also meant that I would not need to get any RGB software working on Linux.

Some online sources suggest disabling the AMD Cool & Quiet function to get extra gaming performance, but I consider this advice to be out-dated, unless you’re overclocking. Instead, I am leaving this setting enabled, then automatically setting the CPU to ‘performance’ mode while gaming via software.

Once I got everything working, I disabled everything that I wasn’t using, and enabled Ultra Fast Boot. This makes the system start faster, but also removes the ability to re-configure the BIOS, because the keyboard will not work. The settings can be reset via a jumper on the motherboard if I need to get back into it.

Software setup

I installed Ubuntu 20.04 LTS. I would suggest sticking with the Long Term Support version of Ubuntu if you are using it for gaming, since Steam is available in the package manager. Bluetooth, Wi-Fi, sound and 3D-accelerated graphics all worked out of the box.

Configuring Ubuntu

These settings make Ubuntu act less like a desktop, and more like a Home Theater PC.

First I enabled automatic login, then cleared the password for the login keyring. This avoids needing to use a keyboard on startup, or getting prompted for a password when web browsers attempt to access the keyring.

Under power settings, I set “Blank Screen” to “Never”, and Automatic Suspend to “Off”.

Under screen lock settings, I set “Blank Screen Delay” to “Never”, and disabled “Automatic Screen Lock”.

Next, I set the theme to dark mode, the display output to 1920×1080 60Hz, and the audio output device to HDMI. The display is 4K, but compatibility will be challenging enough without adding scaling issues in there, so I’m using a lower resolution for now.

Lastly, I set up a tool called psensor to start on boot, so that I could check temperatures.

Installing applications

Linux gaming has advanced a fair bit recently, and since it is a large topic, I won’t try to cover in too much detail here. I’m connecting this computer up to a TV, so this type of setup is only suitable for games with controller support. Still, this includes:

  • Native Linux games.
  • Windows games running via a compatibility layer (WINE/DXVK or Proton).
  • Games for other systems played through an emulator.

An extensive catalogue is available through the Linux version of Steam, which I installed via the Ubuntu package manager. I could have set Steam to launch on startup and call it a day, but there are two other launchers which I installed alongside it:

  • Lutris (from the lutris-team/lutris PPA).
  • Retroarch (from the libretro/testing PPA)

Next, I installed WINE (32 bit and 64 bit), winetricks, and gamemode from the Ubuntu package manager, then Proton GE from GitHub. Proton GE is a widely-used Proton fork, which integrates fixes from upstream WINE. This gave me several different run-times for Windows applications, and simply switching between them or adding some environment variables has been sufficient to work around every compatibility issue that I’ve encountered so-far.

Lastly, I installed Kodi from the team-xbmc/ppa PPA.

Game testing

For this section, I’m listing out a few games from my library that I’ve tested. As a quick reminder, this is all running on integrated graphics, on Linux.

I was mainly checking that I could get a (subjectively) playable frame-rate, and that I could get two controllers to work in local multiplayer where available.

Steam

GRIP combat racing runs via Proton. It is a fast-paced racing game, has local split-screen multiplayer, and runs very smoothly at 720p.

Untitled Goose Game also runs via Proton. It has recently added a 2-player mode, and runs at 1080p with no hiccups.

The Tomb Raider reboot is available native for Linux. With low settings, it runs fine at 1080p, with the benchmark indicating a 99 FPS average.

Shadow of the Tomb Raider, also available native for Linux, is a real challenge for this integrated GPU. I am running it at 1080p, lowest settings, though the benchmark comes back with just a 43 FPS average.

Lutris

Lutris has an Epic Games store installer, which runs the store-front under WINE. I decided to try it out, though it is easily the most buggy software mentioned in this blog post.

The only the only game I tested from there was Rocket League. Native Linux support was recently dropped from this title on Steam, but the Windows version from Epic runs just fine on integrated graphics, and has local multiplayer. I run it at 1080p.

I also tested the original Crysis by manually setting up a WINE prefix and adding it to Lutris. I run this at 720p, medium settings. After a few loops, the benchmark indicates that this averages 112 FPS, so I could probably increase the quality or resolution.

I also tested Crysis 2 via WINE, which had some concerns about my graphics card. I also run it at 720p, using the ‘Gamer’ profile. The benchmark indicates that this averages 60 FPS.

Super Tux Kart is an open source racing game, which can be installed via Lutris. This is a native Linux build, and runs just fine at 1080p. It has controller support and local multiplayer.

Wrap-up

This blog post touches on quite a few topics, so I’ve left out a lot of the details. If you’ve read this far, though, then it’s time to talk about down-sides. Some things did not work as planned.

  • Reading/controlling fan speeds did not work from Linux on this Gigabyte motherboard. This has worked so consistently for me on other hardware, that I did not even think to check for compatibility here. This means I can only set fan speeds in the BIOS.
  • The power supply is noisier than I expected under load. This is due to coil whine, not fan noise.
  • When using the 8BitDo SF30 controller over Bluetooth, RetroArch would pause for 10 or 15 seconds at a time. I tracked this down to the fact that I am connecting it as an Xbox controller, and RetroArch was attempting (and failing) to check its battery level. Connecting the controller after a core has launched avoids the problem.
  • I had hoped to run everything under Wayland, but the Epic Games store had terrible graphical glitches. Everything else on this page (Steam, Lutris, Retroarch) worked on Wayland, and this is apparently due to the way that this app uses OpenGL.

This setup works very well for me, and I am glad to be able to show that it’s possible to do some basic gaming without Microsoft Windows or a dedicated GPU. Still, there are a lot of trade-offs that come from this form-factor and platform, and that’s not for everybody.

If you’re using Linux or a Ryzen APU for gaming, or have tried building in the InWin Chopin case, then please feel free to leave a comment below. I would be interested to know about anything you’ve done differently, and how it worked out.

How to auto-scale the display in GNOME Boxes

I recently installed a virtual machine in GNOME Boxes, and the display was stuck at 1024×768.

The type of display used here is called SPICE, and it includes a channel for auto-scaling. The guest simply needs the agent to be installed.

In this case, I’m running a Debian guest, which means that I must have forgotten to install the spice-vdagent package.

# apt-get install spice-vdagent
Reading package lists... Done
Building dependency tree       
Reading state information... Done
The following NEW packages will be installed:
  spice-vdagent
0 upgraded, 1 newly installed, 0 to remove and 0 not upgraded.
Need to get 47.6 kB of archives.
After this operation, 174 kB of additional disk space will be used.
Get:1 http://deb.debian.org/debian buster/main amd64 spice-vdagent amd64 0.18.0-1 [47.6 kB]
Fetched 47.6 kB in 0s (99.9 kB/s)  
Selecting previously unselected package spice-vdagent.
(Reading database ... 132855 files and directories currently installed.)
Preparing to unpack .../spice-vdagent_0.18.0-1_amd64.deb ...
Unpacking spice-vdagent (0.18.0-1) ...
Setting up spice-vdagent (0.18.0-1) ...
Created symlink /etc/systemd/system/sockets.target.wants/spice-vdagentd.socket → /lib/systemd/system/spice-vdagentd.socket.
[spice-vdagentd.conf:2] Line references path below legacy directory /var/run/, updating /var/run/spice-vdagentd → /run/spice-vdagentd; please update the tmpfiles.d/ drop-in file accordingly.
Processing triggers for man-db (2.8.5-2) ...
Processing triggers for systemd (241-5) ...

The easiest way to ensure that everything is running correctly is to reboot, since the agent will start on boot, and this also forces a new log-in, and a new connection to the display.

# reboot

Result

Assuming that you are otherwise on the default settings, the display in the guest VM will now automatically adjust as you resize the window.

If the “Share Clipboard” setting is enabled for the virtual machine, then spice-vdagent will also enable you to copy & paste text between the host and guest.

How to use the qemu-bridge-helper on Debian 10

If you use the libvirt virtualisation libraries, then you will be familiar with the “user session”. This feature lets you provision virtual machines to run under a regular, unprivileged user account.

The user session is used by GNOME Boxes, and can also be managed from Virtual Machine Manager.

The main downside to this setup is that a regular user can only access a very limited range of networking options. The last time that I mentioned this in a blog post, a reader pointed out that you can actually use qemu-bridge-helper to provide bridged networking to unprivileged virtual machines.

Today I finally tried this out, and it worked really well. With a bit of configuration, you can extend proper networking to this type of VM.

The host

I’m running a graphical Debian 10 desktop, with a few basic virtualisation packages.

  • gnome-boxes for creating VM’s as a local user. This depends on libvirt-daemon, which is enough to host VM’s on the system.
  • virt-manager for a more advanced graphical interface.

The tool that I’m writing about today is qemu-bridge-helper, which is in the qemu-system-common package.

After installation, you will also need to ensure that libvirtd is running.

$ systemctl enable libvirtd.service
$ systemctl start libvirtd.service

Set up a bridge

Libvirt ships with a basic network bridge configuration, you just need to enable it.

Command-line method

Start the default network bridge, and configure it to run on startup.

$ sudo virsh net-autostart --network default
$ sudo virsh net-start --network default

Once this is set up, you should see the bridge virbr0, reporting the IP range 192.168.122.1/24.

$ ip addr show virbr0
3: virbr0: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc noqueue state DOWN group default qlen 1000
    link/ether xx:xx:xx:xx:xx:xx brd ff:ff:ff:ff:ff:ff
    inet 192.168.122.1/24 brd 192.168.122.255 scope global virbr0
       valid_lft forever preferred_lft forever

Graphical method

First, open up Virtual Machine Manager, and authenticate. Right click on QEMU/KVM, and select Details.

Under Virtual NetworksdefaultAutostart, check On Boot, then click Apply.

Setting up qemu-bridge-helper

Create the file /etc/qemu/bridge.conf with the content:

allow virbr0

Restrict the permissions of this file to make sure it can’t be edited by regular users.

# chown root:root /etc/qemu/bridge.conf
# chmod 0640 /etc/qemu/bridge.conf

Add setuid to the qemu-bridge-helper binary.

# chmod u+s /usr/lib/qemu/qemu-bridge-helper

If you do not correctly set this last step, then you will receive the following error when you attempt to connect a VM to the bridge:

Error starting domain: internal error: /usr/lib/qemu/qemu-bridge-helper --use-vnet --br=virbr0 --fd=28: failed to communicate with bridge helper: Transport endpoint is not connected
stderr=failed to create tun device: Operation not permitted

Setting up the VM

Create a virtual machine, either though GNOME Boxes or Virtual Machine Manager. I am using a CentOS VM as an example here, but the guest platform is not particularly important.

Using Virtual Machine Manage, change the network card to the “shared network” virbr0.

The graphical configuration above is equivalent to the following libvirt domain XML, as below.

<interface type='bridge'>
  <mac address='52:54:00:08:5a:7c'/>
  <source bridge='virbr0'/>
  <model type='virtio'/>
  <address type='pci' domain='0x0000' bus='0x00' slot='0x03' function='0x0'/>
</interface>

Result

After restarting the network interface in the guest, I was able to ping the the guest from the host and vice-versa.

This is a significant improvement from “user-mode” networking, which does not facilitate host-to-guest and guest-to-guest communication.

The default virbr0 bridge uses an internal subnet, so the guest here is still inaccessible from the wider LAN. If this doesn’t match your setup, then you can use the same technique to connect unprivileged virtual machines to another bridge of your choice.

Further reading

I had to adapt some paths, user accounts and package names to get this working on Debian. The sources I used are:

Benchmarking PHP code with PhpBench

This blog post is all about measuring the speed of PHP code through micro-benchmarks.

Why micro-benchmarks

I think of micro-benchmarks as a complement to tests. A well-written test would give a developer a definite pass or fail result, while well-written micro-benchmark will give the developer an indication of the duration of an operation.

To put it another way, tests can help you find out whether the code is behaving correctly, while micro-benchmarks can help you to track whether your changes are making the code faster or slower.

Make sure that your code is doing real processing work before you spend time on this: Algorithms such as sorting, compressing, or parsing are good candidates. My guess is that most of the code in a regular PHP application would not benefit significantly from having a suite of benchmarks, because web apps tend to be I/O bound (eg database, network and disk access).

Introducing PhpBench

I have previously written a standalone test script every time I’ve needed to measure the speed of something in PHP. This works up to a point, but it gets quite hard to maintain as the number of benchmarks increases.

PhpBench is one of the available tools for running micro-benchmarks from PHP, and there are a few reasons why I think it’s worth a try:

  • it’s actively maintained
  • it installs with composer
  • it runs in a similar way to PHPUnit, so the benchmarks are fully specified in PHP code and run with a PHP tool.
  • it implements familiar concepts that you would find in benchmarking tools from other languages (eg. JMH from the Java world).

Installation

Start out with a blank composer project.

$ composer init

Add some auto-loading settings to composer.yml that PHP knows to find ExampleApp classes in the src folder.

{
    "name": "mike42/php-benchmark-examples",
    "description": "Example PHP benchmark project",
    "type": "project",
    "license": "MIT",
    "minimum-stability": "dev",
    "require": {},
    "require-dev": {},
    "autoload": {
        "psr-4": {
            "ExampleApp\\": "src"
        }
    }
}

At this point, install the phpbench dev version.

$ composer require --dev phpbench/phpbench:@dev
./composer.json has been updated
Loading composer repositories with package information
Updating dependencies (including require-dev)
Package operations: 22 installs, 0 updates, 0 removals
  - Installing symfony/process (4.4.x-dev 1a42849): Cloning 1a42849a7f from cache
  - Installing symfony/options-resolver (4.4.x-dev 94cbb72): Cloning 94cbb72bb9 from cache
  - Installing symfony/finder (4.4.x-dev 1b5ec12): Cloning 1b5ec12340 from cache
  - Installing symfony/polyfill-ctype (dev-master 82ebae0): Cloning 82ebae0220 from cache
  - Installing symfony/filesystem (4.4.x-dev 5914824): Cloning 59148241f7 from cache
  - Installing psr/log (dev-master c4421fc): Cloning c4421fcac1 from cache
  - Installing symfony/debug (4.4.x-dev 8278839): Cloning 8278839457 from cache
  - Installing symfony/service-contracts (dev-master 0c81a04): Cloning 0c81a04f68 from cache
  - Installing symfony/polyfill-php73 (dev-master d1fb4ab): Cloning d1fb4abcc0 from cache
  - Installing symfony/polyfill-mbstring (dev-master fe5e94c): Cloning fe5e94c604 from cache
  - Installing symfony/console (4.4.x-dev e2fe100): Cloning e2fe1002fd from cache
  - Installing lstrojny/functional-php (1.9.0): Loading from cache
  - Installing beberlei/assert (v3.x-dev ce139b6): Cloning ce139b6bf8 from cache
  - Installing seld/jsonlint (1.7.1): Loading from cache
  - Installing psr/container (dev-master 014d250): Cloning 014d250dae from cache
  - Installing phpbench/container (1.2): Loading from cache
  - Installing webmozart/assert (1.4.0): Loading from cache
  - Installing webmozart/path-util (dev-master 95a8f7a): Cloning 95a8f7ad15 from cache
  - Installing phpbench/dom (0.2.0): Loading from cache
  - Installing doctrine/lexer (dev-master ee614dd): Cloning ee614dd93a from cache
  - Installing doctrine/annotations (1.7.x-dev 3f35255): Cloning 3f35255290 from cache
  - Installing phpbench/phpbench (dev-master dccc67d): Cloning dccc67dd52 from cache
symfony/service-contracts suggests installing symfony/service-implementation
symfony/console suggests installing symfony/event-dispatcher
symfony/console suggests installing symfony/lock
Writing lock file
Generating autoload files

Writing a benchmark

Add some code to src/ExampleThing.php to measure.

<?php

namespace ExampleApp;

class ExampleThing {
  /**
   * Inefficiently multiply numbers together
   */
  public function multiply(int $x, int $y) : int {
    $ret = 0;
    for($i = 0; $i < $x; $i++) {
      for($j = 0; $j < $y; $j++) {
        $ret++;
      }
    }
    return $ret;
  }
}

Write a benchmark for the code in benchmarks/ExampleThingBenchmark.php

<?php

use ExampleApp\ExampleThing;

/**
 * @BeforeMethods({"init"})
 * @Revs(1000)
 * @Iterations(5)
 */
class ExampleThingBenchmark {

    private static $exampleThing;

    public function init()
    {
        self::$exampleThing = new ExampleThing();
    }

    /**
     * @Subject
     */
    public function doMultiply()
    {
        self::$exampleThing -> multiply(100, 100);
    }
}

Before you run anything, you will also need a configuration file. Add this minimal configuration to phpbench.json.dist.

{
    "php_disable_ini": true,
    "bootstrap": "vendor/autoload.php",
    "path": "benchmark",
    "php_config": {
        "extension": [ "json.so" ]
    },
    "time_unit": "milliseconds"
}

Running the benchmark

The most basic way to run all benchmarks at once is phpbench run, which looks like this:

$ php vendor/bin/phpbench run
PhpBench @git_tag@. Running benchmarks.
Using configuration file: /home/mike/workspace/blog/phpbench/php-benchmarks/phpbench.json.dist

\ExampleThingBenchmark

    doMultiply..............................I4 [μ Mo]/r: 0.232 0.232 (ms) [μSD μRSD]/r: 0.000ms 0.10%

1 subjects, 5 iterations, 1,000 revs, 0 rejects, 0 failures, 0 warnings
(best [mean mode] worst) = 0.232 [0.232 0.232] 0.233 (ms)
⅀T: 1.162ms μSD/r 0.000ms μRSD/r: 0.100%

As you might expect, there are options to run specific benchmarks, or to present the results in different ways. However, there were some PhpBench-specific things which you will need to navigate to use it successfully.

Firstly, there are two different output settings to consider:

  • report options allow you to decide which data to print in a table.
  • output options allow you to decide how to format it for output

I also found it useful to add --progress=none to suppress the text displayed by default, since you will otherwise get broken HTML output.

Lastly, be aware that confusingly, the default report is not active by default. Specify it to get a table listing out each iteration.

$ php vendor/bin/phpbench run --progress=none --report=default
suite: 13415431dc0db41c12decee0fa83c26d1db0f678, date: 2019-05-31, stime: 22:01:34
+-----------------------+------------+-----+------+------+----------+-----------+--------------+----------------+
| benchmark             | subject    | set | revs | iter | mem_peak | time_rev  | comp_z_value | comp_deviation |
+-----------------------+------------+-----+------+------+----------+-----------+--------------+----------------+
| ExampleThingBenchmark | doMultiply | 0   | 1000 | 0    | 915,736b | 236.716μs | +1.68σ       | +1.08%         |
| ExampleThingBenchmark | doMultiply | 0   | 1000 | 1    | 915,736b | 231.999μs | -1.45σ       | -0.93%         |
| ExampleThingBenchmark | doMultiply | 0   | 1000 | 2    | 915,736b | 233.960μs | -0.15σ       | -0.1%          |
| ExampleThingBenchmark | doMultiply | 0   | 1000 | 3    | 915,736b | 233.878μs | -0.2σ        | -0.13%         |
| ExampleThingBenchmark | doMultiply | 0   | 1000 | 4    | 915,736b | 234.361μs | +0.12σ       | +0.08%         |
+-----------------------+------------+-----+------+------+----------+-----------+--------------+----------------+

The type of report that I found most useful is aggregate, since it provides a summary of each iteration.

$ php vendor/bin/phpbench run --progress=none --report=aggregate
suite: 1341543c02ebee4031d165665c2724c379bf9c98, date: 2019-05-31, stime: 22:01:18
+-----------------------+------------+-----+------+-----+----------+-----------+-----------+-----------+-----------+---------+--------+-------+
| benchmark             | subject    | set | revs | its | mem_peak | best      | mean      | mode      | worst     | stdev   | rstdev | diff  |
+-----------------------+------------+-----+------+-----+----------+-----------+-----------+-----------+-----------+---------+--------+-------+
| ExampleThingBenchmark | doMultiply | 0   | 1000 | 5   | 915,736b | 231.907μs | 232.614μs | 232.089μs | 233.538μs | 0.713μs | 0.31%  | 1.00x |
+-----------------------+------------+-----+------+-----+----------+-----------+-----------+-----------+-----------+---------+--------+-------+

Both of the example above use console output, which is the default.

HTML reports

To capture phpbench output from a CI environment, you can generate a HTML report as well. Add an output section to phpbench.json.dist.

{
    ...
    "outputs": {
         "html_file": {
             "extends": "html",
             "file": "benchmarks.html",
             "title": "Example benchmark report"
         }
    }
}

This new html_file output can be used with --output=html_file

$ php vendor/bin/phpbench run \
    --progress=none --report=aggregate \
    --output=console --output=html_file

From Jenkins

This is a Jenkinsfile that I use to run benchmarks. You need to install the “HTML Publisher” and “AnsiColor” Jenkins plugins.

pipeline {
  agent any

  stages {
    stage('Install') {
      steps {
        ansiColor('xterm') {
          sh 'composer install'
        }
      }
    }
    stage('Run benchmarks') {
      steps {
        ansiColor('xterm') {
          sh 'php vendor/bin/phpbench run --progress=none --report aggregate --output=console --output=html_file --ansi'
          publishHTML([
            allowMissing: false,
            alwaysLinkToLastBuild: true,
            keepAll: true,
            reportDir: '',
            reportFiles: 'benchmarks.html',
            reportName:
            'phpbench report',
            reportTitles: ''
          ])       
        }
      }
    }
  }
}

Each build is then benchmarked, and the results are available as a link on the left.

The console output of the build shows the results.

To actually view the HTML report you will need to apply this tweak.

Next steps

I’ve started to add phpbench micro-benchmarks to some PHP projects that I work on, and I’m finding it a lot more useful than standalone test scripts.

The next step would be integrate micro-benchmarks better into the development process. For example, I can already review code coverage improvements/regressions of a change on GitHub using Coveralls, which posts a comment to each pull request. This saves you from merging changes which don’t meet your project’s standards.

As far as I can tell, you would need to write custom code for more expressive phpbench reports along these lines, such as:

  • the history of a benchmark over time, or
  • the most-changed benchmarks against a baseline

phpbench does provide the building-blocks though, since you can set some options to archive the results of each run to a folder.

How to integrate Gitea and Jenkins

Gitea is a web app for hosting Git repositories. It’s open source, and very simple to get running. With some extra setup, it can also trigger Jenkins builds, and display the Jenkins build status of each commit once it has been built.

Because the documentation for the Jenkins plugin is very minimalist, I decided to write about it for future reference.

About this setup

I installed Jenkins and Gitea on the same Debian 9 server on the LAN. They communicate only over HTTP, so they could just as easily be installed separately.

To make the configuration clear, I’ve used jenkins.example.com in URLs which refer to Jenkins, and gitea.example.com for the Gitea.

Gitea installation

This command will install and start the linux-amd64 version of Gitea as the user “git”.

useradd git -r --create-home && \
  mkdir /opt/gitea && chown -R git: /opt/gitea && \
  wget -O /opt/gitea/gitea https://dl.gitea.io/gitea/1.7.0/gitea-1.7.0-linux-amd64 && \
  chmod +x /opt/gitea/gitea && \
  sudo -u git bash -c "cd /opt/gitea && ./gitea web"

Shut it down, and configure some paths at /opt/gitea/custom/conf/app.ini. These will depend on your environment.

SSH_DOMAIN       = gitea.example.com
DOMAIN           = gitea.example.com
HTTP_PORT        = 3000
ROOT_URL         = http://gitea.example.com:3000/

Start it back up as a systemd service at this point, by creating /etc/systemd/system/gitea.service with this content:

[Unit]
Description=gitea
After=network.target

[Service]
ExecStart=/opt/gitea/gitea web
WorkingDirectory=/opt/gitea
User=git
Type=simple

[Install]
WantedBy=multi-user.target

Once this is saved, start the service.

systemctl daemon-reload
systemctl start gitea

Optionally, also configure it to start on boot.

systemctl enable gitea

Jenkins installation

I installed Jenkins from the official Debian repo at jenkins-ci.org, and clicked through the initial install.

Add plugin

Open up Manage JenkinsManage Plugins. Navigate to Available and check the Gitea plugin.

Next, install the plugin and restart Jenkins.

The configuration for the plugin is located under Manage JenkinsConfigure System.

At this point you will want to tell Jenkins where to find your Gitea server.

I don’t suggest choosing Manage hooks, because it uses the same account to manage hooks across all repos, which would violate the principle of least privilege.

Set up a project in Gitea

In Gitea, create a project, then a repository under that.

Register an account in Gitea for Jenkins to use for this project.

Log out, log back in as yourself, and add Jenkins as a collaborator to the repo, with Write access.

This is the only permission you need for public repositories. If you plan to lock down your Gitea organization later, then you will also need to give this Jenkins account Read access at the organization level.

Set up a project in Jenkins

Add a new Gitea Organization Jenkins job.

Enter the name of the organization, and the account to log in with.

Add the details for the new account, and make sure it’s selected.

The other options don’t need to be changed at this stage.

When you press ‘save’, Jenkins will immediately attempt to find any repositories in the Gitea organization, and kick off any builds. Unless everything is correct, this is unlikely to work the first time, so pay attention to error logs.

These three places will show what’s happening:

  • Scan Gitea Organization Log, which lists repositories in the organization.
  • Scan Multibranch Pipeline Log for each repository, which shows the discovery of branches.
  • Console output for each build, which will show errors if the build status could not be submitted.

Problems which I’ve found here include:

  • The URL of Gitea in the Jenkins configuration must match the URL to Gitea in its own configuration.
  • The Jenkins user account must have permission to list repositories, clone, and update statuses.
  • Empty repositories, and repositories without a ‘Jenkinsfile’ are ignored.

For that last step, here is an empty Jenkinsfile that you can put in your repository to test this integration:

pipeline {
    agent any

    stages {
        stage('Do nothing') {
            steps {
                sh '/bin/true'
            }
        }
    }
}

Once this is sorted out, you can expect to see your repository in Jenkins.

Every branch with a Jenkinsfile will appear.

And each time a commit is mentioned in Gitea, it will display a small icon to indicate the build status.

Set up a web hook in Gitea

At this point, builds need to be manually triggered. To trigger them each time the repository changes, we need to get a notification out to Jenkins.

Under the repository settings, click WebhooksAdd webhookGitea.

The correct values to use are:

  • URL: http://[ your jenkins server ]/gitea-webhook/post
  • POST Content Type: application/json

Once you press Add Webhook, the path will appear with a small grey dot, indicating that it hasn’t been run before.

If you click edit, then the Test Delivery button can be used to check that it’s working.

The icon indicates the status. If things aren’t working correctly, then click the delivery UUID to expand the full request information, which should help with debugging.

Final result

With Jenkins and Gitea, you have a simple self-hosted a continuous integration environment.

In Gitea, you can store, update and review your code. Any build and test steps in a Jenkinsfile will be run automatically each time the repository changes.

The detailed output for each build is visible in Jenkins, where you can track build results with a variety of plugins.

Handling I/O errors in PHP

This blog post is all about how to handle errors from the PHP file_get_contents function, and others which work like it.

The file_get_contents function will read the contents of a file into a string. For example:

<?php
$text = file_get_contents("hello.txt");
echo $text;

You can try this out on the command-line like so:

$ echo "hello" > hello.txt
$ php test.php 
hello

This function is widely used, but I’ve observed that error handling around it is often not quite right. I’ve fixed a few bugs involving incorrect I/O error handling recently, so here are my thoughts on how it should be done.

How file_get_contents fails

For legacy reasons, this function does not throw an exception when something goes wrong. Instead, it will both log a warning, and return false.

<?php
$filename = "not-a-real-file.txt";
$text = file_get_contents($filename);
echo $text;

Which looks like this when you run it:

$ php test.php 
PHP Warning:  file_get_contents(not-a-real-file.txt): failed to open stream: No such file or directory in test.php on line 3
PHP Stack trace:
PHP   1. {main}() test.php:0
PHP   2. file_get_contents() test.php:3

Warnings are not very useful on their own, because the code will continue on without the correct data.

Error handling in four steps

If anything goes wrong when you are reading a file, your code should be throwing some type of Exception which describes the problem. This allows developers to put a try {} catch {} around it, and avoids nasty surprises where invalid data is used later.

Step 1: Detect that the file was not read

Any call to file_get_contents should be immediately followed by a check for that false return value. This is how you know that there is a problem.

<?php
$filename = "not-a-real-file.txt";
$text = file_get_contents($filename);
if($text === false) {
  throw new Exception("File was not loaded");
}
echo $text;

This now gives both a warning and an uncaught exception:

$ php test.php 
PHP Warning:  file_get_contents(not-a-real-file.txt): failed to open stream: No such file or directory in test.php on line 3
PHP Stack trace:
PHP   1. {main}() test.php:0
PHP   2. file_get_contents() test.php:3
PHP Fatal error:  Uncaught Exception: File was not loaded in test.php:5
Stack trace:
#0 {main}
  thrown in test.php on line 5

Step 2: Suppress the warning

Warnings are usually harmless, but there are several good reasons to suppress them:

  • It ensures that you are not depending on a global error handler (or the absence of one) for correct behaviour.
  • The warning might appear in the middle of the output, depending on php.ini.
  • Warnings can produce a lot of noise in the logs

Use @ to silence any warnings from a function call.

<?php
$filename = "not-a-real-file.txt";
$text = @file_get_contents($filename);
if($text === false) {
  throw new Exception("File was not loaded");
}
echo $text;

The output is now only the uncaught Exception:

$ php test.php 
PHP Fatal error:  Uncaught Exception: File was not loaded in test.php:5
Stack trace:
#0 {main}
  thrown in test.php on line 5

Step 3: Get the reason for the failure

Unfortunately, we lost the “No such file or directory” message, which is pretty important information, which should go in the Exception. This information is retrieved from the old-style error_get_last method.

This function might just return empty data, so you should check that everything is set and non-empty before you try to use it.

<?php
$filename = "not-a-real-file.txt";
error_clear_last();
$text = @file_get_contents($filename);
if($text === false) {
  $e = error_get_last();
  $error = (isset($e) && isset($e['message']) && $e['message'] != "") ?
      $e['message'] : "Check that the file exists and can be read.";
  throw new Exception("File '$filename' was not loaded. $error");
}
echo $text;

This now embeds the failure reason directly in the message.

$ php test.php 
PHP Fatal error:  Uncaught Exception: File 'not-a-real-file.txt' was not loaded. file_get_contents(not-a-real-file.txt): failed to open stream: No such file or directory in test.php:9
Stack trace:
#0 {main}
  thrown in test.php on line 9

Step 4: Add a fallback

The last time I introduced error_clear_last()/get_last_error() into a code-base, I learned out that HHVM does not have these functions.

Call to undefined function error_clear_last()

The fix for this is to write some wrapper code, to verify that each function exists.

<?php
$filename = 'not-a-real-file';
clearLastError();
$text = @file_get_contents($filename);
if ($text === false) {
    $error = getLastErrorOrDefault("Check that the file exists and can be read.");
    throw new \Exception("Could not retrieve image data from '$filename'. $error");
}
echo $text;

/**
 * Call error_clear_last() if it exists. This is dependent on which PHP runtime is used.
 */
function clearLastError()
{
    if (function_exists('error_clear_last')) {
        error_clear_last();
    }
}
/**
 * Retrieve the message from error_get_last() if possible. This is very useful for debugging, but it will not
 * always exist or return anything useful.
 */
function getLastErrorOrDefault(string $default)
{
    if (function_exists('error_clear_last')) {
        $e = error_get_last();
        if (isset($e) && isset($e['message']) && $e['message'] != "") {
            return $e['message'];
        }
    }
    return $default;
}

This does the same thing as before, but without breaking other PHP runtimes.

$ php test.php 
PHP Fatal error:  Uncaught Exception: Could not retrieve image data from 'not-a-real-file'. file_get_contents(not-a-real-file): failed to open stream: No such file or directory in test.php:7
Stack trace:
#0 {main}
  thrown in test.php on line 7

Since HHVM is dropping support for PHP, I expect that this last step will soon become unnecessary.

How not to handle errors

Some applications put a series of checks before each I/O operation, and then simply perform the operation with no error checking. An example of this would be:

<?php
$filename = "not-a-real-file.txt";
// Check everything!
if(!file_exists($filename)) {
  throw new Exception("$filename does not exist");
}
if(!is_file($filename)) {
  throw new Exception("$filename is not a file");
}
if(!is_readable($filename)) {
  throw new Exception("$filename cannot be read");
}
// Assume that nothing can possibly go wrong..
$text = @file_get_contents($filename);
echo $text;

You could probably make a reasonable-sounding argument that checks are a good idea, but I consider them to be misguided:

  • If you skip any actual error handling, then your code is going to fail in more surprising ways when you encounter an I/O problem that could not be detected.
  • If you do perform correct error handling as well, then the extra checks add nothing other than more branches to test.

Lastly, beware of false positives. For example, the above snippet will reject HTTP URL’s, which are perfectly valid for file_get_contents.

Conclusion

Most PHP code now uses try/catch/finally blocks to handle problems, but the ecosystem really values backwards compatibility, so existing functions are rarely changed.

The style of error reporting used in these I/O functions is by now a legacy quirk, and should be wrapped to consistently throw a useful Exception.

The top 100 Ansible modules

Ansible is an automation system which is widely used for deployments and configuration. It contains a dizzying array of modules for interfacing with things like files, services, package managers, and various pieces of software and hardware.

Background

There are some modules which I would expect every Ansible user to be familiar with, while others are barely used at all. I could have a guess about which modules I use most, but what about the wider community?

I couldn’t find any existing data, so I set out to determine which Ansible modules were most widely-used. The best source for this is Ansible Galaxy, which is a directory of around 18,000 Ansible roles. Using some scripts, I sifted through the 18,259 publicly accessible roles from Ansible Galaxy, found 71,610 files to look through, and tallied up which modules were being used by each of the 317,757 tasks in those files.

The list

These are the top 100 modules as of January 2019. I used the same method to build such a list in January 2018, so you can also see the change in popularity of some modules over the last 12 months.

Rank Module Uses Rank +/-
1 file 26,224 +1
2 include 25,267 -1
3 template 24,062
4 command 23,952
5 service 19,436
6 shell 19,401
7 set_fact 17,079 +1
8 apt 16,006 +1
9 lineinfile 12,432 -2
10 copy 8,794 +1
11 yum 8,733 -1
12 assert 7,504 +7
13 include_tasks 6,686 +15
14 stat 5,922 -2
15 package 5,422 -1
16 get_url 5,323 -3
17 debug 5,147 -2
18 import_tasks 5,016 +20
19 include_vars 4,251 -2
20 apt_repository 3,789 -4
21 user 3,233 -3
22 fail 2,812 +2
23 unarchive 2,734 -2
24 apt_key 2,665 -4
25 pip 2,474
26 systemd 2,389 +8
27 action 2,114 -4
28 git 2,043 -1
29 uri 1,680 +7
30 group 1,665 -1
31 sysctl 1,331 -5
32 raw 1,323 -2
33 mysql_user 1,266
34 meta 1,204 +3
35 replace 1,125 -3
36 ini_file 1,125 -5
37 find 1,030 -15
38 local_action 986 -3
39 mysql_db 959
40 cron 925
41 wait_for 898
42 rpm_key 771
43 include_role 742 +15
44 yum_repository 723 +2
45 mount 669 -2
46 blockinfile 619 +3
47 firewalld 579 +1
48 ufw 563 -1
49 authorized_key 557 -4
50 docker_container 488 +2
51 dnf 455 +5
52 seboolean 417 -8
53 homebrew 409 -2
54 fetch 378 +1
55 npm 367 -5
56 osx_defaults 363 +8
57 postgresql_user 350 -4
58 pkgng 345 -4
59 pause 340 +4
60 script 314 +5
61 setup 290 +7
62 postgresql_db 289 -2
63 mysql_replication 286 -1
64 win_regedit 281 +6
65 pacman 257 -4
66 debconf 256 +1
67 slurp 255 +10
68 gem 253 -11
69 iptables 240 +13
70 apache2_module 231 -4
71 synchronize 213 -2
72 docker 213 -13
73 alternatives 210
74 selinux 202
75 oc_obj 199 +98
76 make 194 -1
77 win_shell 191 +13
78 modprobe 187 +11
79 hostname 181 -7
80 zypper 174 -4
81 xml 160
82 supervisorctl 160 -11
83 win_file 149 +15
84 homebrew_cask 149 -6
85 add_host 144 -2
86 rabbitmq_user 129 -7
87 pamd 124 +81
88 win_command 116 +13
89 assemble 116 -9
90 htpasswd 115 -6
91 apk 112 -5
92 openbsd_pkg 111 -7
93 win_get_url 109 +14
94 win_chocolatey 109
95 docker_image 109 +4
96 tempfile 106 +25
97 locale_gen 105 -5
98 easy_install 97 -10
99 django_manage 97 -3
100 composer 96 -3

Out of over 2,500 modules mentioned in Ansible Galaxy, it turns out that only a few dozen are widely used. Following this, there is a long tail of infrequently-used and custom modules.

How many modules do I need to know?

You can write 80% of the roles in Ansible Galaxy using only the 74 most popular Ansible modules.

For writing an individual Ansible role, you don’t need anywhere near this many. The median number of modules used by an role in Ansible Galaxy is just 6.

There are a few reasons that you won’t need to use many popular modules all at the same time:

  • Different technologies: eg. mysql_db and postgresql_db set up different databases, and would be invoked by different roles if you need both.
  • Obsolete modules: eg. docker has been split up, and new code will use docker_container instead.
  • Different levels of abstraction: eg. cross-platform developers will use the package module, while develpers who know their platform may directly use the apt, dnf and yum modules. You are unlikely to see these mixed.
  • Different styles: include_tasks vs. import_tasks.

Some notes

Here are some things to keep in mind when reading the table.

  • I skipped any role that was not hosted in a public GitHub repository, or could not be parsed. A php-yaml bug also kept a few files out of the 2019 tally.
  • Each task was counted only once, so local_action and action are obscuring which modules were eventually executed.
  • The OpenShift oc_obj module on this list is not bundled with Ansible (the module itself is distributed through Galaxy).
  • I did not attempt to exclude abandoned or incomplete packages. Eg. The docker module on this list has been removed from the latest version, but remains popular in existing Ansible Galaxy roles.

Thermal Sans Mono: A new bitmap font for receipt printers

Today I’m writing about a new project that I’ve started to create a font, called “Thermal Sans Mono”. I’m hoping to make this into a set of specialized free bitmap fonts for use on thermal receipt printers and printer emulators.

Here is what it looks like so far:

In this blog post I’ll talk a bit about why I needed a new font, how it’s derived from GNU Unifont, and what I’m planning to do with it.

Background

Firstly, I am solving a technical problem, not a design problem.

Printing non-ASCII characters on receipts can get a bit complex, especially if you have a lot of different printers to support. Each thermal receipt printer has a different set of available “code pages” of glyphs, which makes localisation a real adventure.

Using the escpos-printer-db project, we can currently create software which accepts UTF-8 text, and switches between the available code pages to render text for a specific printer. I’m now looking at how we can encode characters that are not not in any available code page.

For example, in this code from the python-escpos library, the ? subsitution character still has to be printed for some characters:

encoding = self.encoder.find_suitable_encoding(text[0])
if not encoding:
    self._handle_character_failed(text[0])

Rather than send a ? character, I think we could retrieve a glyph from a suitable font and send it as a bitmap instead. I previously wrote about a technique to cache glyphs from GNU Unifont in printer memory, but these glyphs don’t have the correct size or weight to be displayed alongside the glyphs that the printer has rendered itself.

Of course we can’t do this yet, because there is no suitable bitmap font.

How it’s made

The closest font to what we need is GNU Unifont, which is thankfully freely licensed. A Unifont glyph is typically 8×16 with a thin line and a lot of whitespace, while I need to print in 12×24, with thick lines and no whitespace.

To control as many variables as possible, I decided to pre-render a transformed version of GNU Unifont. This involved tracing the original glyph and re-drawing it with new parameters.

The first step is to use a purpose-built algorithm to trace the lines from the glyph:

The points are then re-drawn onto a new canvas with a thicker outline, and a lot less whitespace.

When this canvas is scaled back to 12×24, we have our glyph.

This glyph is now the correct size and weight for use on most thermal receipt printers.

Roadmap

Currently, I have only processed the ASCII characters. As a GNU Unifont derivative, we should (with some effort) be able to include many more glyphs in the future.

As a starting goal, I am aiming to make the font complete enough to include as a fallback font in the escpos-php printing library.

Download

I’ve made my initial work work on “Thermal Sans Mono” available as a bitmap font PCF and BDF format on GitHub. You can find these in the “Releases” section of the GitHub project.

View on GitHub →

As always, bugs, comments and suggestions are appreciated.