Category Archives: Code

Whitelisting an email or domain in Mailgun

Leave a reply

Recently a client who uses Mailgun to send email from their website had a problem where one of the clients own email addresses had been “suppressed” in Mailgun, meaning Mailgun would no longer send email to that address. It was caused by the client’s mail server having a problem for long enough that Mailgun decided they couldn’t send email to the particular address.

Mailgun have the ability to “whitelist” individual email addresses or whole domains so they will not stop sending to them. That’s exactly what I needed to put in a long term fix for this problem. It’s not available through their website, but you can do it through their API. That was fine, we’re using the API for other things so I can adapt some code to use it for this.

Unfortunately, the docs for whitelisting an email address are currently incorrect. I had some back and forth with their support people and they gave me the following curl example for how to whitelist an email address:

curl -i -X POST  -u api:[your api key] https://api.eu.mailgun.net/v3/[your mailgun domain]/whitelists -F address=[your email address]

Where [your api key] and [your mailgun domain] and [your email address] get replaced by your details.

Remember that if you’re in Europe, you’re probably using the EU API, not the one in the docs, so you’ll need to change the API domain to https://api.eu.mailgun.net

To whitelist a whole domain, use:

curl -i -X POST  -u api:[your api key] https://api.mailgun.net/v3/[your mailgun domain]/whitelists -F domain=[your email domain]

Finally, to check the whitelist, use this:

curl -i -X GET  -u api:[your api key]  https://api.mailgun.net/v3/[your mailgun domain]/whitelists

This was a frustrating situation as it took a while for their support to understand that giving me back the incorrect examples from the docs wasn’t helping, but they did come through for us in the end and were polite and trying to be helpful throughout. Hopefully they’ll get their docs updated soon.

Using ACF Blocks by default in a new post in WordPress

Leave a reply

One of my clients has a website using Advanced Custom Fields (ACF) Blocks to style everything in the site, including their News section.

To make a new post in News, they had to load a Pattern of the relevant Blocks. This being a multi-step process, and a bit annoying if you just want to get on with posting your article, I looked for a way to load the blocks into the new post by default.

You can do this by putting this code in your functions.php (with amends explained below)

function site_post_register_template() {
    $post_type_object = get_post_type_object( 'post' );
    $post_type_object->template = array(
        array( 'my-blocks/news-header' ),
        array( 'my-blocks/news-article' ),
        array( 'my-blocks/news-article-related' ),
    );
}
add_action( 'init', 'site_post_register_template' );

In this case, I have three Blocks loading – the specific header, the article, and a strip of related articles. I can load in as many Blocks as I want.

To make this work for you, find where you have saved your blocks and list them out in the same way, in an array for each.

You will have to change ‘my-blocks’ to the prefix of your ACF Blocks. To find that prefix, go to the directory where you store the code for the relevant Block and open the block.json file. Copy the details from the “name” line to be inside the array( '' ), above.

Now, when a new Post is started, it automatically loads in the ACF Blocks and my client can get on with posting their news, not thinking about how they load in the Pattern and where they left my instructions for doing so.

Fixing a Mailgun API unknown domain error

Leave a reply

I’m using the Mailgun API for a couple of clients, making sure we don’t keep sending email to someone who has marked their previous message as spam.

It should be quite simple to use as the docs are very clear, but I kept getting an ‘unknown domain’ error in the returned message when I used the API with one of my client’s domains rather than the sandbox domain the Mailgun provides.

The fix was to use the EU address for the API: api.eu.mailgun.net rather than the standard api.mailgun.net. As my clients are in the UK, they are put into the EU servers, rather than the American ones.

I didn’t find a simple suggestion to do that, so I’m writing this so I find it next time this trips me up.

Fixing minor issues moving code from Railo to Lucee ColdFusion

Leave a reply

I have a couple of personal projects at a host that offers the ColdFusion web programming language. I have a soft spot for ColdFusion as its my first programming language and I still find it very easy to put together sites in it as it comes with easy to use functionality the other server side language I use a lot, PHP, only gets through frameworks where they’ve been added by others.

Adobe, current owners of the official ColdFusion engine, charge rather a lot for licenses and that has a knock on effect on the price of hosting. So, I use Viviotech, who offer the open source engines that run the ColdFusion language. The popular one of those was Railo and is now Lucee. Viviotech recently shifted my sites onto Lucee, which I’d been meaning to try as Railo is old and now discontinued.

I hit a couple of problems:

Gotcha 1: No CGI.PATH_INFO

In the project, I redirect all requests through a single routing file. Within that file, I was using CGI.PATH_INFO to get the path of the page being requested. That stopped working, which turns out to be because the host is using Nginx and that doesn’t have PATH_INFO turned on by default. There are ways of making it work, but I didn’t want to be doing support requests to do that and it may not have been accepted for my cheap-as-chips hosting package.

Instead, I make the redirect in the .htaccess file send the path through for me.

My redirect went from this:
RewriteRule ^(.*)$ ./routing.cfm/$1 [L]
To this:
RewriteRule ^(.*)$ /routing.cfm?path=$1 [NC,L,QSA]
Which gives me the URL of the page being requested in URL.path (apart from the / that is normally at the start, which I needed in part of my URL detection, so I added it on.)

I re-wrote my code to use the new URL.path (with added /) instead of CGI.PATH_INFO and that got it working again.

Gotcha 2: Saving files needs permissions set

In one of the sites, I get an image from an API that makes screenshots, and save it locally so I don’t have to use the API over and over. That means getting the image using CFHTTP, then saving it using CFFILE.

That worked, but I couldn’t open the files. The fix was to use MODE=”644″ within CFFILE. This set the file permissions so the image file can be read by the world, and show up on a web page.

<cffile
action = "write"
file = "<path><filename>"
output = "#cfhttp.fileContent#"
mode="644">

Improvement: Can read SSL protected RSS feeds

Railo couldn’t read RSS feeds that were protected by Let’s Encrypt SSL certificates and some of the other cheap/free SSL providers. Lucee can.

That’s great, as I made a very basic proxy (not really worthy of the word) to go request the SSL protected RSS feed through a PHP script I had on some other hosting, which would then send it through without SSL. Not great for security (although these are all public posts it is reading.) So the update to Lucee let me remove the ‘proxy’, which has simplified my code and maintenance.

Now I have my sites working again, I’m looking forward to delving into Lucee some more.

Fixing Vagrant VMs after a VirtualBox upgrade

Leave a reply

I use virtual machines to split up my web development projects. This lets me have a small Linux server run inside my Mac, allowing me to match the operating system with what the code will run in on a public service, without going all the way to running Linux all the time. As I want an easy life when it comes to setting these up, I use Vagrant to make and manage the virtual machines within VirtualBox, which runs them for me.

Recently I set up an extra Vagrant virtual machine (VM) to hold two projects for a client and keep them separate from some of my own projects, as they needed particular settings within Linux and a different database server to run them than the settings and database for my own projects.

I used a Homestead VM, as I knew the settings were good. Homestead is made for Laravel PHP sites, but is perfectly valid for lots of other PHP projects. My client uses various versions of the Symfony framework, and they work fine within Homestead built boxes.

The new VM worked fine but the existing VM stopped working. I could start it, I could SSH into it, but I couldn’t get any websites from it to show in the browser. Which is a big problem, as that’s all it is there for.

After much faff and investigation, I discovered the problem is a bug in the current version of VirtualBox. I had updated it while setting up the new VM, but unlike previous upgrades, this caused a problem. VirtualBox needs VMs to use IP addresses within the 192.168.56.0/21 range or the “bridge” between MacOS and the virtual machine doesn’t work, so no web pages show.

The IP of the old Homestead box my own projects were in was 192.168.10.10, which was the default when I installed it. That used to work, but now does not. The new VM uses 192.168.56.4, which is within the allowed range, so it worked fine.

To fix the old VM:

I had to edit the Homestead.yaml file. At the top where it says:

ip: "192.168.10.10"

I changed it to:

ip: "192.168.56.10"

I then ran:

vagrant reload --provision

To get vagrant to reconfigure the existing Homestead box to use this changed IP address, without deleting anything inside the VM.

And finally I edited the entries in my hosts file (which is in /etc for me in MacOS) for the websites in the VM, changing their IP from 192.168.10.10 to 192.168.56.10

Once all that was done, the websites in the VM started working again.

This was a very, very frustrating problem. I spent a long time investigating what was happening inside the VM as I presumed that’s where the problem was, eventually stumbling on the solution after searching for wider and wider versions of the problem. Thanks to Tom Westrick whose post on Github got me to the solution.