So I've run an upgrade on a 2.4 site and all is kind-of well after migrating the usual deprecated classes and template constructs. Other than these changes, nothing in my classes or templates has changed.

What I have found on doing the upgrade on two different environments is that some weird encoding is making it's way into some of my page content. Live example is here: http://ss3.vivaexpeditions.com/south-america-tours/patagonia-travel/wild-patagonia-santiago-to-ushuaia-2. See the  and – characters in the main body.

Admittedly, these characters are in the database too. But the thing is, they are also in the SS2.4 database too, and that somehow gets away with it. http://www.vivaexpeditions.com/south-america-tours/patagonia-travel/wild-patagonia-santiago-to-ushuaia-2

Database encoding looks identical. Doctype and meta looks identical. When I compare sources in Winmerge, only the content really stands out as being different.

Could this be something new in SS3 at work that I need to cater for? Anyone encountered the same? I've made a quick function to clean on the fly, but would rather they overhead if there are other ideas.

Cheers

I have since discovered that this is occurring on most content after the upgrade.

After checking the database collations against the 2.4 copy of the site, I can safely say the only difference here is the Silverstripe version. The 2.4 version even has content with these encoded characters in them, yet they are not displayed on render.

Has no one else seen this? Days of hunting general encoding issues has revealed nothing.