We have the following site configuration for Simplified Chinese:

title: 'Simplified Chinese'
enabled: true
base: /简/
typo3Language: cn
locale: zh_CN.UTF-8
iso-639-1: zh
navigationTitle: 简
hreflang: zh-Hans
direction: ltr
fallbackType: fallback
fallbacks: '0'
flag: cn
languageId: '4'
websiteTitle: ''

This configuration leads to a 404 error.

What happens¶

When a user navigates to a url including 简 as a language param, the path is automatically url encoded by the browser or the server (we use Nginx). In the `Request` object the sign becomes `%E7%AE%80`.

Middleware SiteResolver¶

This middleware tries to match the request url with the site configuration to get the correct site and language for the request.

SiteMatcher¶

It delegates the matching to `SiteMatcher::matchRequest()`. Within `SiteMatcher::getRouteCollectionForAllSites()` the site configuration is loaded from YAML and a `\TYPO3\CMS\Core\Routing\Route` is created from the routing settings. The `$path´ variable holds the decoded Chinese sign.

UrlMatcher¶

The method returns a collection of routes built from the site configuration, which gets passed to `Symfony\Component\Routing\Matcher\UrlMatcher::__construct()`. Right after that, `SiteMatcher` delegates the matching to `Symfony\Component\Routing\Matcher::match($pathinfo)`, with the current path from the `Request` object.

When passed to `match()` the value is `%E7%AE%80` but this gets passed deeper down to `Symfony\Component\Routing\Matcher\UrlMatcher::matchCollection()` but before the path gets decoded by `rawurldecode()`:

if ($ret = $this->matchCollection(rawurldecode($pathinfo) ?: '/', $this->routes)) {
    return $ret;
}

Inside `Symfony\Component\Routing\Matcher\UrlMatcher::matchCollection()` finally the matching happens by comparing `$trimmedPathinfo` with `$staticPrefix` or in other words the encoded path `%E7%AE%80` with the decoded path 简.

Patch¶

My patch resolves the problem by decoding the base path from site configuration before it gets stored into `\TYPO3\CMS\Core\Routing\Route`.