Mailing List Articles Atom Feed Comments Atom Feed Twitter Reddit Facebook

Tag Cloud

3d 3d printing account algorithms android announcement architecture archives arduino artificial intelligence artix assembly async audio automation backups bash batch blender blog bookmarklet booting bug hunting c sharp c++ challenge chrome os cluster code codepen coding conundrums coding conundrums evolved command line compilers compiling compression containerisation css dailyprogrammer data analysis debugging demystification distributed computing dns docker documentation downtime electronics email embedded systems encryption es6 features ethics event experiment external first impressions freeside future game github github gist gitlab graphics hardware hardware meetup holiday holidays html html5 html5 canvas infrastructure interfaces internet interoperability io.js jabber jam javascript js bin labs learning library linux lora low level lua maintenance manjaro minetest network networking nibriboard node.js open source operating systems optimisation own your code pepperminty wiki performance phd photos php pixelbot portable privacy problem solving programming problems project projects prolog protocol protocols pseudo 3d python reddit redis reference releases rendering resource review rust searching secrets security series list server software sorting source code control statistics storage svg systemquery talks technical terminal textures thoughts three thing game three.js tool tutorial tutorials twitter ubuntu university update updates upgrade version control virtual reality virtualisation visual web website windows windows 10 worldeditadditions xmpp xslt

Avoiding accidental array mutation when iterating arrays in PHP

Pepperminty Wiki is written in PHP, and I've posted before about the search engine I've implemented for it that's powered by an inverted index. In this post, I want to talk about an anti-feature of PHP that doesn't behave the way you'd expect, and how to avoid running into the same problem I did.

To do this, let's introduce a simple example of the problem at work:

$arr = [];
for($i = 0; $i < 3; $i++) {
    $key = random_int(0, 2000);
    $arr[$key] = $i;
    echo("[init] key: $key, i: $i\n");

foreach($arr as $key => &$value) {
    // noop

echo("structure before: "); var_dump($arr);

foreach($arr as $key => $value) {
    echo("key: $key, i was $value\n");

echo("structure after: "); var_dump($arr);

The above code initialises an associative array with 3 elements. The contents might look like this:

Key Value
469 0
1777 1
1685 2

Pretty simple so far. It then iterates over it twice: once referring to the values by reference (that's what the & there is for), and the second time referring to the items by value.

You'd expect the array to be identical before and after the second foreach loop, but you'd be wrong:

Key Value
469 0
1777 1
1685 1

Wait, what? That's very odd. What's going on here? How can a foreach loop that's iterating an array by value mutate an array? To understand why, let's take a step back for a moment. Here's another snippet:


$arr = [ 1, 2, 3 ];

foreach($arr as $key => $value) {
    echo("$key: $value\n");

echo("The last value was $key: $value\n");

What do you expect to happen here? While in Javascript with a for..of loop with a let declaration both $key and $value would have fallen out of scope by now, in PHP foreach statements don't create a new scope for variables. Instead, they inherit the scope from their parent - e.g. the global scope in the above or their containing function if defined inside a function.

To this end, we can still access the values of both $key and $value in the above example even after the foreach loop has exited! Unexpected.

It gets better. Try prefixing $value with an ampersand & in the above example and re-running it - note that both $key and $value are both still defined.

This leads us to why the unexpected behaviour occurs. For some reason because of the way that PHP's foreach loop is implemented, if we re-use the same variable name for $value here in a subsequent loop it replaces the value of the last item in the array.

Shockingly enough this is actually documented behaviour (see also this bug report), though I'm somewhat confused as to how it happens on the last element in the array instead of the first.

With this in mind, to avoid this problem in future if you iterate an array by reference with a foreach loop, always remember to unset() the $value, like so:

$arr = [];
for($i = 0; $i < 3; $i++) {
    $key = random_int(0, 2000);
    $arr[$key] = $i;
    echo("[init] key: $key, i: $i\n");

foreach($arr as $key => &$value) {
    // noop
unset($key); unset($value);

echo("structure before: "); var_dump($arr);

foreach($arr as $key => $value) {
    echo("key: $key, i was $value\n");

echo("structure after: "); var_dump($arr);

By doing this, you can ensure that you don't accidentally mutate your arrays and spend weeks searching for the bug like I did.

It's language features like these that catch developers out: and being aware of the hows and whys of their occurrence can help you to avoid them next time (if anyone can explain why it's the last element in the array that's affected instead of the first, I'd love to know!).

Regardless, although I'm aware of how challenging implementing a programming language is, programming language designers should take care to avoid unexpected behaviour like this that developers don't expect.

Found this interesting? Comment below!

Sources and further reading

Disassembling .NET Assemblies with Mono

As part of the Component-Based Architectures module on my University course, I've been looking at what makes the .NET ecosystem tick, and how .NET assemblies (i.e. .NET .exe / .dll files) are put together. In the process, we looked as disassembling .NET assemblies into the text-form of the Common Intermediate Language (CIL) that they contain. The instructions on how to do this were windows-specific though - so I thought I'd post about the process on Linux and other platforms here.

Our tool of choice will be Mono - but before we get to that we'll need something to disassemble. Here's a good candidate for the role:

using System;

namespace SBRL.Demo.Disassembly {
    static class Program {
        public static void Main(string[] args) {
            int a = int.Parse(Console.ReadLine()), b = 10;
                "{0} + {1} = {2}",
                a, b,
                a + b

Excellent. Let's compile it:

csc Program.cs

This should create a new Program.exe file in the current directory. Before we get to disassembling it, it's worth mentioning how the compilation and execution process works in .NET. It's best explained with the aid of a diagram:

Left-to-right flowchart: Multiple source languages get compiled into Common Intermediate Language, which is then executed by an execution environment.

As is depicted in the diagram above, source code in multiple languages get compiled (maybe not with the same compiler, of course) into Common Intermediate Language, or CIL. This CIL is then executed in an Execution Environment - which is usually a virtual machine (Nope! not as in Virtual Box and KVM. It's not a separate operating system as such, rather than a layer of abstraction), which may (or may not) decide to compile the CIL down into native code through a process called JIT (Just-In-Time compilation).

It's also worth mentioning here that the CIL code generated by the compiler is in binary form, as this take up less space and is (much) faster for the computer to operate on. After all, CIL is designed to be efficient for a computer to understand - not people!

We can make it more readable by disassembling it into it's textual equivalent. Doing so with Mono is actually quite simple:

monodis Program.exe >

Here I redirect the output to a file called for convenience, as my editor has a plugin for syntax-highlighting CIL. For those reading without access to Mono, here's what I got when disassembling the above program:

.assembly extern mscorlib
  .ver 4:0:0:0
  .publickeytoken = (B7 7A 5C 56 19 34 E0 89 ) // .z\V.4..
.assembly 'Program'
  .custom instance void class [mscorlib]System.Runtime.CompilerServices.CompilationRelaxationsAttribute::'.ctor'(int32) =  (01 00 08 00 00 00 00 00 ) // ........

  .custom instance void class [mscorlib]System.Runtime.CompilerServices.RuntimeCompatibilityAttribute::'.ctor'() =  (
        01 00 01 00 54 02 16 57 72 61 70 4E 6F 6E 45 78   // ....T..WrapNonEx
        63 65 70 74 69 6F 6E 54 68 72 6F 77 73 01       ) // ceptionThrows.

  .custom instance void class [mscorlib]System.Diagnostics.DebuggableAttribute::'.ctor'(valuetype [mscorlib]System.Diagnostics.DebuggableAttribute/DebuggingModes) =  (01 00 07 01 00 00 00 00 ) // ........

  .hash algorithm 0x00008004
  .ver  0:0:0:0
.module Program.exe // GUID = {D6162DAD-AD98-45B3-814F-C646C6DD7998}

.namespace SBRL.Demo.Disassembly
  .class private auto ansi beforefieldinit Program
    extends [mscorlib]System.Object

    // method line 1
    .method public static hidebysig 
           default void Main (string[] args)  cil managed 
        // Method begins at RVA 0x2050
    // Code size 47 (0x2f)
    .maxstack 5
    .locals init (
        int32   V_0,
        int32   V_1)
    IL_0000:  nop 
    IL_0001:  call string class [mscorlib]System.Console::ReadLine()
    IL_0006:  call int32 int32::Parse(string)
    IL_000b:  stloc.0 
    IL_000c:  ldc.i4.s 0x0a
    IL_000e:  stloc.1 
    IL_000f:  ldstr "{0} + {1} = {2}"
    IL_0014:  ldloc.0 
    IL_0015:  box [mscorlib]System.Int32
    IL_001a:  ldloc.1 
    IL_001b:  box [mscorlib]System.Int32
    IL_0020:  ldloc.0 
    IL_0021:  ldloc.1 
    IL_0022:  add 
    IL_0023:  box [mscorlib]System.Int32
    IL_0028:  call void class [mscorlib]System.Console::WriteLine(string, object, object, object)
    IL_002d:  nop 
    IL_002e:  ret 
    } // end of method Program::Main

    // method line 2
    .method public hidebysig specialname rtspecialname 
           instance default void '.ctor' ()  cil managed 
        // Method begins at RVA 0x208b
    // Code size 8 (0x8)
    .maxstack 8
    IL_0000:  ldarg.0 
    IL_0001:  call instance void object::'.ctor'()
    IL_0006:  nop 
    IL_0007:  ret 
    } // end of method Program::.ctor

  } // end of class SBRL.Demo.Disassembly.Program

Very interesting. There are a few things of note here:

  • The metadata at the top of the CIL tells the execution environment a bunch of useful things about the assembly, such as the version number, the classes contained within (and their signatures), and a bunch of other random attributes.
  • An extra .ctor method has been generator for us automatically. It's the class' constructor, and it automagically calls the base constructor of the object class, since all classes are descended from object.
  • The ints a and b are boxed before being passed to Console.WriteLine. Exactly what this does and why is quite complicated, and best explained by this Stackoverflow answer.
  • We can deduce that CIL is a stack-based language form the add instruction, as it has no arguments.

I'd recommend that you explore this on your own with your own test programs. Try changing things and see what happens!

  • Try making the Program class static
  • Try refactoring the int.Parse(Console.ReadLine()) into it's own method. How is the variable returned?

This isn't all, though. We can also recompile the CIL back into an assembly with the ilasm code:


This makes for some additional fun experiments:

  • See if you can find where b's value is defined, and change it
  • What happens if you alter the Console.WriteLine() format string so that it becomes invalid?
  • Can you get ilasm to reassemble an executable into a .dll library file?

Found this interesting? Discovered something cool? Comment below!

Sources and Further Reading

Help! My SQLite database is malformed!

Recently I came across a rather worrying SQLite database error:

Error: database disk image is malformed

Hrm, that's odd. Upon double-checking, it looked like the database was functioning (mostly) fine. The above error popping up randomly was annoying though, so I resolved to do something about it. Firstly, I double-checked that said database was actually 'corrupt':

sudo sqlite3 path/to/database.sqlite 'PRAGMA integrity_check';

This outputted something like this:

*** in database main ***
Main freelist: 1 of 8 pages missing from overflow list starting at 36376
Page 23119: btreeInitPage() returns error code 11
On tree page 27327 cell 30: 2nd reference to page 27252

Uh oh. Time to do something about it then! Looking it up online, it turns out that the 'best' solution out there is to export to an .sql file and then reimport again into a fresh database. That's actually quite easy to do. Firstly, let's export the existing database to an .sql file. This is done via the following SQL commands (use sqlite3 path/to/database.db to bring up a shell)

.mode insert
.output /tmp/database_dump.sql

With the database exported, we can now re-import it into a fresh database. Bring up a new SQLite3 shell with sqlite3, and do the following:

.save /tmp/new_database.sqlite
.read /tmp/database_dump.sql

...that might take a while. Once done, swap our old corrupt database out for your shiny new one and you're done! I would recommend keeping the old one as a backup for a while just in case though (perhaps bzip2 path/to/old_database.sqlite?).

Also, if the database is on an embedded system, you may find that downloading it to your local computer for the repair process will make it considerably faster.

Found this useful? Still having issues? Comment below!


Debug your systemd services with journalctl

Some nice bulbs and bits of wood.

The chances are that if you're using linux, you will probably have run into systemd. If you find yourself in the situation where you've got a systemd service that keeps dying and you don't know why (I've been there before several times!), and there's nothing helpful in /var/log, before you give up, you might want to give journalctl a try. It's systemd's way of capturing the output of a service and storing it in it's logging system (or something).

When I first found out about it, I read that apparently journalctl -xe servicename would show me the logs for any given service. It turned out that it wasn't the case (it just threw a nasty error), so I went trawling through the man pages and found the correct command-line switch. If you've got a service called rocketbooster.service, and you want to see if systemd has any logs stored for it, then you can execute this command:

journalctl --unit rocketbooster.service

...or for short

journalctl -u rocketbooster.service

It should open the logs (if there are any) in less - with the oldest logs at the top, so you might need to scroll all the way down to the bottom to see anything that's relevant to your problem (shift + G will take you to the bottom of the file).

I've found that systemd has a habit of rotating the logs too - and journalctl doesn't appear to know how to access the rotated logs, so it's best if you use this command as soon as possible after failure (suggestions on how to access these rotated logs are welcome! Post down in the comment :D).

I thought I'd document it here in case it was useful to anyone - and so I don't forget myself! :P

Profiling PHP with XDebug

(This post is a fork of a draft version of a tutorial / guide originally written as an internal document whilst at my internship.)

The PHP and xdebug logos.

Since I've been looking into xdebug's profiling function recently, I've just been tasked with writing up a guide on how to set it up and use it, from start to finish - and I thought I'd share it here.

While I've written about xdebug before in my An easier way to debug PHP post, I didn't end up covering the profiling function - I had difficulty getting it to work properly. I've managed to get it working now - this post documents how I did it. While this is written for a standard Debian server, the instructions can easily be applied to other servers.

For the uninitiated, xdebug is an extension to PHP that aids in the debugging of PHP code. It consists of 2 parts: The php extension on the server, and a client built into your editor. With these 2 parts, you can create breakpoints, step through code and more - though these functions are not the focus of this post.

To start off, you need to install xdebug. SSH into your web server with a sudo-capable account (or just use root, though that's bad practice!), and run the following command:

sudo apt install php-debug

Windows users will need to download it from here and put it in their PHP extension direction. Users of other linux distributions and windows may need to enable xdebug in their php.ini file manually (windows users will need extension=xdebug.dll; linux systems use instead).

Once done, xdebug should be loaded and working correctly. You can verify this by looking the php information page. To see this page, put the following in a php file and request it in your browser:


If it's been enabled correctly, you should see something like this somewhere on the resulting page:

Part of the php information page, showing the xdebug section.

With xdebug setup, we can now begin configuring it. Xdebug gets configured in php.ini, PHP's main configuration file. Under Virtualmin each user has their own php.ini because PHP is loaded via CGI, and it's usually located at ~/etc/php.ini. To find it on your system, check the php information page as described above - there should be a row with the name "Loaded Configuration File":

The 'loaded configuration file' directive on the php information page.

Once you've located your php.ini file, open it in your favourite editor (or type sensible-editor php.ini if you want to edit over SSH), and put something like this at the bottom:



Obviously, you'll want to customise the above. The xdebug.profiler_enable_trigger_value directive defines a secret key we'll use later to turn profiling on. If nothing else, make sure you change this! Profiling slows everything down a lot, and could easily bring your whole server down if this secret key falls into the wrong hands (that said, simply having xdebug loaded in the first place slows things down too, even if you're not using it - so you may want to set up a separate server for development work that has xdebug installed if you haven't already). If you're not sure on what to set it to, here's a bit of bash I used to generate my random password:

dd if=/dev/urandom bs=8 count=4 status=none | base64 |  tr -d '=' | tr '+/' '-_'

The xdebug.profiler_output_dir lets you change the folder that xdebug saves the profiling output files to - make sure that the folder you specify here is writable by the user that PHP is executing as. If you've got a lot of profiling to do, you may want to consider changing the output filename, since xdebug uses a rather unhelpful filename by default. The property you want to change here is xdebug.profiler_output_name - and it supports a number of special % substitutions, which are documented here. I can recommend something phpprofile.%t-%u.%p-%H.%R.%cachegrind - it includes a timestamp and the request uri for identification purposes, while still sorting chronologically. Remember that xdebug will overwrite the output file if you don't include something that differentiates it from request to request!

With the configuration done, we can now move on to actually profiling something :D This is actually quite simple. Simply add the XDEBUG_PROFILE GET (or POST!) parameter to the url that you want to test in your browser. Here are some examples:


Adding this parameter to a request will cause xdebug to profile that request, and spit out a cachegrind file according to the settings we configured above. This file can then be analysed in your favourite editor, or, if it doesn't have support, an external program like qcachegrind (Windows) or kcachegrind (Everyone else).

If you need to profile just a single AJAX request or similar, most browsers' developer tools let you copy a request as a curl or wget command (Chromium-based browsers, Firefox - has an 'edit and resend' option), allowing you to resend the request with the XDEBUG_PROFILE GET parameter.

If you need to profile everything - including all subrequests (only those that pass through PHP, of course) - then you can set the XDEBUG_PROFILE parameter as a cookie instead, and it will cause profiling to be enabled for everything on the domain you set it on. Here's a bookmarklet that set the cookie:

javascript:(function(){document.cookie='XDEBUG_PROFILE='+'insert_secret_key_here'+';expires=Mon, 05 Jul 2100 00:00:00 GMT;path=/;';})();


Replace insert_secret_key_here with the secret key you created for the xdebug.profiler_enable_trigger_value property in your php.ini file above, create a new bookmark in your browser, paste it in (making sure that your browser doesn't auto-remove the javascript: at the beginning), and then click on it when you want to enable profiling.

Was this helpful? Got any questions? Let me know in the comments below!

Sources and further reading

An easier way to debug PHP

A nice view of some trees taken by Mythdael I think, the awesome person who designed this website :D

Recently at my internship I've been writing quite a bit of PHP. The language itself is OK (I mean it does the job), but it's beginning to feel like a relic of a bygone era - especially when it comes to debugging. Up until recently I've been stuck with using echo() and var_dump() calls all over the place in order to figure out what's going on in my code - that's the equivalent of debugging your C♯ ACW with Console.WriteLine() O.o

Thankfully, whilst looking for an alternative, I found xdebug. Xdebug is like visual studio's debugging tools for C♯ (albeit a more primitive form). They allow you to add breakpoints and step though your PHP code one line at a time - inspecting the contents of variables in both the local and global scope as you go. It improves the standard error messages generated by PHP, too - adding stack traces and colour to the output in order to make it much more readable.

Best of all, I found a plugin for my primary web development editor atom. It's got some simple (ish) instructions on how to set up xdebug too - it didn't take me long to figure out how to put it to use.

I'll assume you've got PHP and Nginx already installed and configured, but this tutorial looks good (just skip over the MySQL section) if you haven't yet got it installed. This should work for other web servers and configurations too, but make sure you know where your php.ini lives.

XDebug consists of 2 components: The PHP extension for the server, and the client that's built into your editor. Firstly, you need to install the server extension. I've recorded an asciicast (terminal recording) to demonstrate the process:

(Above: An asciinema recording demonstrating how to install xdebug. Can't see it? Try viewing it on

After that's done, you should be able to simply install the client for your editor (I use php-debug for atom personally), add a breakpoint, load a php page in your web browser, and start debugging!

If you're having trouble, make sure that your server can talk directly to your local development machine. If you're sitting behind any routers or firewalls, make sure they're configured to allow traffic though on port 9000 and configured to forward it on to your machine.

TeleConsole: A simple remote debugging solution for awkward situations

Header image (Header image from here.)

Several times in the last few months I've found myself in some kind of awkward situation where I need to debug a C♯ program, but the program in question either doesn't have a console, or is on a remote machine. In an ideal world, I'd like to have all my debugging message sent to my development machine for inspection. That way I don't have to check the console of multiple different machines just to get an idea as to what has gone wrong.

C♯ already has System.Diagnostics.Debug, which functions similarly to the Console class, except that it sends data to the Application output window. This is brilliant for things running on your local machine through Visual Studio or MonoDevelop, but not so great when you've got a program that utilises the network and has to run on separate computers. Visual Studio for one starts to freeze up if you open the exact same repository on a network drive that's already open somewhere else.

It is for these reasons that I finally decided to sit down and write TeleConsole. It's a simple remote console that you can use in any project you like. It's split into 2 parts: the client library and the server binary. The server runs on your development machine, and listens for connections from the client library, which you reference in your latest and greatest project and use to send messages to the server.

Take a look here: sbrl/TeleConsole (GitLab) (Direct link to the releases page)

The client API is fully documented with intellisense comments, so I assume it should be very easy to work out how to use it (if there's something I need to do to let you use the intellisense comments when writing your own programs, let me know!). If you need some code to look at, I've created an example whilst testing it.

Although it's certainly not done yet, I'll be definitely be building upon it in the future to improve it and add additional features.

Test C♯ code online with

I've known about for a while now. It is a site that provides you with a REPL (Read-Eval-Print-Loop) for many different languages, without you having to install the language in question thanks to the native client.

A REPL (in case you didn't know) is like a command prompt, but for a specific programming language or environment. For example, if you type node into your command prompt (if you have Node.js installed), it will start a REPL for you to play around with.

Recently I have discovered that also supports C♯ (via the mono compiler version at the time of typing), and it lets you write, compile and run C♯ code without ever leaving your browser. I was so surprised by this I thought that I'd make a blog post about it. Apparently you can even embed things you've created into other pages too - here's a small test program I wrote whilst playing around with it:


Update: Corrected The expansion of REPL.

Prolog Visualisation Tool

A visualisation of depth first search in Prolog.

Recently, I've been finding that Prolog is getting rather more complicated, and that the traces that I keep doing are getting longer and longer. This is making it rather difficult to understand what's going on, and so in response to this I am building the Prolog Visualisation Tool(kit).

Basically, the Prolog Visualisation Tool(kit) is a tool that, given a Prolog trace, produces a diagram of the trace in question. The image at the top of this post is diagram produced by the tool for a depth first search.

You can find it live now on GitHub Pages.

It is built with mermaid, a really cool diagramming library by knsv, which converts some custom graph syntax to an svg.

The next step will be to animate it, but I haven't got that far yet. Expect an update soon!

Reading HTTP 1.1 requests from a real web server in C#

I've received rather a lot of questions recently asking the same question, so I thought that I 'd write a blog post on it. Here's the question:

Why does my network client fail to connect when it is using HTTP/1.1?

I encountered this same problem, and after half an hour of debugging I found the problem: It wasn't failing to connect at all, rather it was failing to read the response from the server. Consider the following program:

using System;
using System.IO;
using System.Net.Sockets;

class Program
    static void Main(string[] args)
        TcpClient client = new TcpClient("", 80);
        client.SendTimeout = 3000;
        client.ReceiveTimeout = 3000;
        StreamWriter writer = new StreamWriter(client.GetStream());
        StreamReader reader = new StreamReader(client.GetStream());
        writer.WriteLine("GET /path HTTP/1.1");

        string response = reader.ReadToEnd();

        Console.WriteLine("Got Response: '{0}'", response);

If you change the hostname and request path, and then compile and run it, you (might) get the following error:

An unhandled exception of type 'System.IO.IOException' occurred in System.dll

Additional information: Unable to read data from the transport connection: A connection attempt failed because the connected party did not properly respond after a period of time or established connection failed because connected host has failed to respond.

Strange. I'm sure that we sent the request. Let's try reading the response line by line:

string response = string.Empty;
    string nextLine = reader.ReadLine();
    response += nextLine;
    Console.WriteLine("> {0}", nextLine);
} while (reader.Peek() != -1);

Here's some example output from my server:

> HTTP/1.1 200 OK
> Server: nginx/1.9.10
> Date: Tue, 09 Feb 2016 15:48:31 GMT
> Content-Type: text/html
> Transfer-Encoding: chunked
> Connection: keep-alive
> Vary: Accept-Encoding
> strict-transport-security: max-age=31536000;
> 2ef
> <html>
> <head><title>Index of /libraries/</title></head>
> <body bgcolor="white">
> <h1>Index of /libraries/</h1><hr><pre><a href="../">../</a>
> <a href="prism-grammars/">prism-grammars/</a>
   09-Feb-2016 13:56                   -
> <a href="blazy.js">blazy.js</a>                                           09-F
eb-2016 13:38                9750
> <a href="prism.css">prism.css</a>                                          09-
Feb-2016 13:58               11937
> <a href="prism.js">prism.js</a>                                           09-F
eb-2016 13:58               35218
> <a href="smoothscroll.js">smoothscroll.js</a>
   20-Apr-2015 17:01                3240
> </pre><hr></body>
> </html>
> 0

...but we still get the same error. Why? The reason is that the web server is keeping the connection open, just in case we want to send another request. While this would usually be helpful (say in the case of a web browser - it will probably want to download some images or something after receiving the initial response), it's rather a nuisance for us, since we don't want to send another request and it's rather awkward to detect the end of the response without detecting the end of the stream (that's what the while (reader.Peek() != -1); is for in the example above).

Thankfully, there are a few solutions to this. Firstly, the web server will sometimes (but not always - take the example response above for starters) send a content-length header. This header will tell you how many bytes follow after the double newline (\r\n\r\n) that separate the response headers from the response body. We could use this to detect the end of the message. This is the reccommended way , according to RFC2616.

Another way to cheat here is to send the connection: close header. This instructs the web server to close the connection after sending the message (Note that this will break some of the tests in the ACW, so don't use this method!). Then we can use reader.ReadToEnd() as normal.

A further cheat would be to detect the expected end of the message that we are looking for. For HTML this will practically always be </html>. We can close the connection after we receive this line (although this doesn't work when you're not receiving HTML). This is seriously not a good idea. The HTML could be malformed, and not contain </html>.

Art by Mythdael