Problem with displaying stream titles with diacritic symbols

2024-03-17 23:35:52

Hi,
there is a problem that when the stream has a special letter in the title, e.g. "ó", foobar stops displaying titles before the given letter appears
For example, stream title is
"Kraków"
and is displayed as
"Krak"
In this case it is an Icecast stream, but it is not always active. That's why I'm not providing a link to it yet.

Re: Problem with displaying stream titles with diacritic symbols

Reply #1 – 2024-03-18 01:18:26

There are two issues I am aware of that cause problems here.

#1 You must use Unicode of some form, and you are instead using US ASCII (or Windows ANSI); use UTF-8 encoding instead with a Byte Order Marker preferred.  If it is a playlist you are using, use m3u8 extension with UTF-8 instead of m3u.
#2 If you are running an older release of Windows 10, you have to change the font that supports Unicode characters.  You will face this issue if you see squares for some characters instead of expected fonts (like musical notes, for example)

Re: Problem with displaying stream titles with diacritic symbols

Reply #2 – 2024-03-18 07:59:06

Unfortunately it sounds like the source server is sending the data incorrectly. Old Icecast servers didn't understand Unicode and used whatever codepage the machines were running. This resulted in all non-ASCII characters to change into wrong symbols when the listener didn't use the same codepage on their machine.

Now Icecast uses Unicode with common UTF-8 encoding and should support all languages without problems. Unfortunately the character encoding appears to still be a problem. Icecast documentation mentions under "charset" option that they seem to be well aware of encoding problems on the serving side. The explanation is a bit worrying. All tag standards have been able to support unicode for decades, so any issues there must be from ancient files, buggy software or misconfigurations.

The title getting cut off is likely caused by the title claiming to be UTF-8 encoded but actually isn't. The string decoder probably stops processing the text when it detects illegal characters.

I assume it would be rather simple for Peter to let the string pass through as-is when decoding fails, though that brings us back to the old codepage mismatch issue where the text can be entirely wrong. But my assumption could also be entirely wrong and this problem could be something entirely different.

Re: Problem with displaying stream titles with diacritic symbols

Reply #3 – 2024-03-18 18:00:44

This stream doesn't work now, so I can't show you a live example. But this applies to the Icecast 2.4.4 server
Moreover, in Winamp 5.9.0 Build 9999, the stream title displays correctly.